From in%@vtcs1 Thu Jun 4 09:09:50 1987 Date: Thu, 4 Jun 87 09:09:30 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #138 Status: RO Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 4 Jun 87 03:18 EDT Received: from relay.cs.net by RELAY.CS.NET id aa04545; 4 Jun 87 0:00 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa05460; 3 Jun 87 23:58 EDT Date: Wed 3 Jun 1987 20:28-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #138 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Thursday, 4 Jun 1987 Volume 5 : Issue 138 Today's Topics: Queries - Uncertainty in ART & Expert Systems for Debugging and Porting & Sources for June AI Expert & AAAI at Seattle & Common LISP on PRIME 50 & Connectionist AI Grad Schools, Education - AI Graduate Schools, Philosophy - Computational Complexity, Binding - Walter Bunch, Funding - Travel Funding, Education - Computer Grading and the Law, Theory - Linguistic Precision & Symbol Grounding ---------------------------------------------------------------------- Date: 1 Jun 87 16:02:51 GMT From: ihnp4!alberta!calgary!arcsun!greg@ucbvax.Berkeley.EDU (Greg Sidebottom) Subject: Request for information: uncertainty in ART I am using ART (from INFERENCE) and I am interested in implementing a mechanism for dealing with uncertainty. I would like to hear from anybody who has addressed this problem. Thanks in advance Greg -- Greg Sidebottom, Alberta Research Council 3rd Floor, 6815 8 Street N.E. Calgary, Alberta CANADA T2E 7H7 (403) 297-2677 UUCP: ...!{ubc-vision, alberta}!calgary!arcsun!greg ------------------------------ Date: 3 Jun 87 03:15:49 GMT From: eric@eddie.mit.edu (Eric Van Tassell) Subject: Expert Systems, Debugging and Porting Hi, Does anyone have any experience in building expert systems to assist in porting large C (or any language) programs to new hardware and OS environments? I am interested in building a system to aid in porting and debugging a 100K line relational database and 4GL. Please e-mail to me any info you think might be relevant. (Success, failure, elation, bitter dejection, or utter frustration) Thanks in advance. Eric Van Tassell eric@eddie.mit.edu ------------------------------ Date: 1 Jun 87 20:52:20 GMT From: tektronix!tekcrl!tekchips!stever@ucbvax.Berkeley.EDU (Steve Rehfuss) Subject: sources for June AI Expert Can someone send me the source code posted for the June issue of AI Expert? I actually just want the prolog benchmark stuff, if you happen to have it separated out. Sorry about this, it expired before I knew I wanted it. Thanks, Steve R stever%tekchips.tek.com@relay.cs.net ------------------------------ Date: 3 Jun 87 10:10:00 EST From: "LIZ_FONG" Reply-to: "LIZ_FONG" Subject: Information on AAAI at Seattle Can some one send me info on AAAI at Seattle on July 13-17 E. Fong ------------------------------ Date: 3 Jun 87 18:35:56 GMT From: necntc!primerd!doug@ames.arpa (Douglas Rand) Subject: Common LISP on PRIME 50 Series I'm interested in people's reaction to Lucid's CL on the Prime. Are people even aware that this exists? Doug Rand (...!mit-eddie!primerd!doug, doug@primerd.prime.com) -- Douglas Rand, Prime Computer Inc. (decvax!necntc!primerd!doug) -> The above opinions are mine alone and are not influenced by narcotics, my employer, my cat or the clothes I wear. ------------------------------ Date: 25 May 87 20:00:48 GMT From: speech2.cs.cmu.edu!yamauchi@pt.cs.cmu.edu (Brian Yamauchi) Subject: Connectionist AI Grad Schools I will be graduating from Carnegie-Mellon next May, with a BS in applied math/computer science, and I am planning to attend graduate school with the goal of a PhD in computer science. My field of interest is artificial intelligence, specifically, connectionist artificial intelligence. I am currently consdiering Carnegie-Mellon, MIT, Caltech, Stanford, UCSD, and the University of Rochester. Are there any other universities that I should be considering? Are there any universities conducting connectionist AI research that I have missed? I would greatly appreciate any information that anyone could provide. Also, I would be interested in hearing any opinions about the relative merits of the computer science graduate programs at these universities, both in general and relative to my specific interests. Thanks in advance, Brian Yamauchi Brian Yamauchi ARPANET: yamauchi@speech2.cs.cmu.edu Carnegie-Mellon University Computer Science Department ------------------------------ Date: 2 Jun 87 02:48:13 GMT From: decvax!dartvax!takis@ucbvax.Berkeley.EDU (Takis Metaxas) Subject: Re: Need info on grad schools with a good AI program From my experience, I can point out to you two schools with some projects in AI: Brown Univ. in Prov.,RI has speciality in natural language representation, and Carnegie-Mellon in searching. Good luck with the field you have chosen... [See back issues of AI Magazine and the SIGART Newsletter for descriptions of many graduate programs. -- KIL] ------------------------------ Date: Tue, 2 Jun 87 04:43:53 EDT From: Jim Hendler Subject: Re: philosophy and computational complexity I second the suggestion of the Cherniak paper. If you want a more complete work try CHristopher Cherniak, Minimal Rationaliy I believe it is MIT Press 87. ------------------------------ Date: 1 Jun 87 09:56:12 GMT From: mcvax!ukc!its63b!hwcs!aimmi!walt@seismo.css.gov (Walter Bunch) Subject: Conceptual Information Research When I made my original posting, my .signature address was incorrect. Thanks to those who got their response to me anyway. Our address changed recently. Sorry for the trouble. -- Walter Bunch, Scottish HCI Centre, Ben Line Building, Edinburgh, EH1 1TN UUCP: walt@uk.ac.hw.cs ARPA: walt%cs.hw.ac.uk@cs.ucl.ac.uk JANET: walt@uk.ac.hw.cs "Is that you, Dave?" ------------------------------ Date: Tue, 2 Jun 87 15:24:37 BST From: "G. Joly" (Birkbeck) Subject: Travel Funding. With reference to the articles on information on financial support to travel to Milan for IJCAI-87, the following are possible sources of support. (1) Royal Society (U.K. residents and Ph.D. status only). (2) British Computer Society (members only). (3) AI and Simulation of Behaviour (members only). (4) AAAI (members only?). In the case of (1) above, the passing date has already gone, but the information may be of use in the future. Most professional bodies seem to have some funds available to their own members. I am not 100% sure of all of the above, but hope this short list is a start (does anyone have a larger collection?). Gordon Joly, Computer Science, Birkbeck College, Malet Street, LONDON WC1E 7HX. +44 1 631 6468 ARPA: gjoly@cs.ucl.ac.uk BITNET: UBACW59%uk.ac.bbk.cu@AC.UK UUCP: ...!seismo!mvcax!ukc!bbk-cs!gordon ------------------------------ Date: Wed, 3 Jun 87 10:48:31 PDT From: Neil Hunt Subject: Re: Computer Grading and the Law In V5 #135, Laurence Leff makes a point about computer grading of student essays. He proposes that using computers to grade essays on a first pass, with "some procedure for complaints to be made to a human being with an appropriate hearing" and that the com- puter "must in some way indicate how the grade was deter- mined". I think that he has missed the point of earlier discussions expressing concern over the educational consequences of hav- ing students orient their efforts towards pleasing a machine rather than a human grader. I believe that the real lesson that students would learn in this situation, is that it is much simpler to write their essays in a style that would satisfy the mechanical grader than to pursue rectification of their grades by requesting a hearing with a human. In fact, most students would probably soon discover how to beat the machine at its own game, writing in a style which would be unacceptable to a human grader, but which a machine with rules of a limited scope might grade highly. The opposite side of the coin, however, as most students are aware, is that human graders all have their own preferences and foibles. Students do learn to avoid certain techniques and foster others just because their human graders seem to dislike the former and like that latter, even if these feel- ings are not representative of all graders. The advantage of human involvement is that the scope of the human includes an understanding of this very problem, thereby providing a curb on the possibility of either the teacher or the student exploiting the situation too far. Of course, the problem is a characteristic of our society, as one's work is always judged by people with prejudices and biases. I believe that before we introduce additional com- puterised agents of judgement, we should have a good under- standing of all the problems they might pose. This is not to say that mechanical style checkers do not have their place. Perhaps all students should have the option of using such a tool before submitting their work to the human grader, but they should be encouraged to under- stand its limitations as well as its strengths, and avoid falling into the trap of assuming that if the machine liked their essay, that the intended readership would also like it. Perhaps it is a little premature to be considering the legality of using computerised grading systems. I am sure that there are many legal options available to teachers and graders which we would not expect them to utilise if they were not effective teaching and learning tools. I think that the desirability of using such an option should be established before time is wasted debating whether it is legal. Neil/. These are my own opinions and not those of my employer etc. ------------------------------ Date: 2 Jun 87 00:39:52 GMT From: hoptoad!laura@ucbvax.Berkeley.EDU (Laura Creighton) Reply-to: hoptoad!laura@ucbvax.Berkeley.EDU (Laura Creighton) Subject: Re: framing problems In article <8705280722.AA09419@ucbvax.Berkeley.EDU> DAVIS@EMBL.BITNET writes: > >I'd like to briefly say that perhaps an even more astounding problem >than that proposed by Stevan Harnad is that connected with the means by >which literate, intelligent and interested persons can totally obscure >the central core of an idea by the use of unnecessarily obtuse jargon. > >If we're going to discuss the more philosphical end of AI (*yes please!*), >then we don't *have* to throw everyone else off the track by bogging >down the chat in a maze of terms intended to have *such* a precise meaning >as to prevent anyone but the author from truly grasping the intended meaning. Precision is a good thing. If one can say precisely what one means then one will not be misunderstood. This, alas, is a pipe dream. There is no way to say precisely what one means -- what one says does not have precise meaning embedded in the words or the relationships between the words. Rather, one shares a linguistic context with one's audience. This means that the serach for precision is never ending. Right now there are a good number of people who want to talk about ``psychic energy'' and ``interpersonal energy'' and the like. Reguardless of what these people mean by these terms, it is clear that they do not mean m c-squared. The search for precision continues. -- (C) Copyright 1987 Laura Creighton - you may redistribute only if your recipients may. ``One must pay dearly for immortality: one has to die several times while alive.'' -- Nietzsche Laura Creighton ihnp4!hoptoad!laura utzoo!hoptoad!laura sun!hoptoad!laura ------------------------------ Date: 2 Jun 87 12:54:00 EST From: cugini@icst-ecf.arpa Reply-to: Subject: physical invertibility and symbol grounding S. Harnad writes: > Now I conjecture that it is this physical invertibility -- the possibility > of recovering all the original information -- that may be critical in > cognitive representations. I agree that there may be information loss in > A/A transformations (e.g., smoothing, blurring or loss of some > dimensions of variation), but then the image is simply *not analog in > the properties that have been lost*! It is only an analog of what it > preserves, not what it fails to preserve..... > > A strong motivation for giving invertibility a central role in > cognitive representations has to do with the second stage of A/D > conversion: symbolization. The "symbol grounding problem" that has > been under discussion here concerns the fact that symbol systems > depend for their "meanings" on only one of two possibilities: One is > an interpretation supplied by human users -- "`Squiggle' means `animal' and > `Squoggle' means `has four legs'" -- and the other is a physical, causal > connection with the objects to which the symbols refer. .... > > The reason the invertibility must be physical rather than merely > formal or conceptual is to make sure the system is grounded rather > than hanging by a skyhook from people's mental interpretations. I wonder why the grounding is to depend on invertibility rather than causation and/or resemblance? Isn't it true that physically distinct kinds of light (eg. #1 red-wavelength and green-wavelength vs. #2 yellow-wavelength) can cause completely indistinguishable sensations (ie subjective yellow)? Is this not, then, a non-invertible, but nonetheless grounded sensation? When I experience something as yellow, I have no way short of spectroscopy of knowing what the "real" physical characteristics are of the light. Nonetheless, I know what "yellow" means, as do young children, scientifically naive people, etc. I don't have a ready-made candidate to substitute for invertibility as a basis for symbol-grounding, although I suspect, as mentioned above, that causation and resemblance are lurking around somewhere. But how can invertibility serve if in fact our sensations are, in general, not invertible? John Cugini ------------------------------ Date: Tue, 2 Jun 87 13:48:51 pdt From: ladkin@kestrel.ARPA (Peter Ladkin) Subject: symbol grounding stevan harnad writes guardedly: > perhaps there was an element of incoherence in all but the most > technical and restricted of signal-analytic candidates. for the record, my suggestion was not signal-analytic, and no-one showed any element of technical incoherence. however, it was met with resounding uninterest, since it was a distinction from logic. since most people want an inherent distinction, i.e. one that maintains under translations and coding, my suspicion is still that technical logic and complexity theory, not signal processing, is the place to look for a solution. peter ladkin ladkin@kestrel.arpa ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jun 11 15:49:14 1987 Date: Thu, 11 Jun 87 15:49:06 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #139 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 11 Jun 87 15:47 EDT Received: from relay.cs.net by RELAY.CS.NET id aa04579; 10 Jun 87 3:53 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa06087; 10 Jun 87 3:46 EDT Date: Wed 10 Jun 1987 00:13-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #139 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Wednesday, 10 Jun 1987 Volume 5 : Issue 139 Today's Topics: Conference - AAAI's Preregistration Deadline, Binding - Number Theory Net, Queries - Small Expert Systems & Speech Data Compression & Dominoes & Hofstadter's Waking Up From the Boolean Dream, Theory - Applying AI Models to Biology ---------------------------------------------------------------------- Date: Thu, 4 Jun 87 10:23:18 PDT From: AAAI Subject: AAAI's Preregistration Deadline The AAAI would like to remind those individuals interested in attending AAAI-87 in Seattle, July 13-17, that the preregistration deadline of Friday, June 12, draws very near. If you would like registration materials, please call or send us a msg with your name and mailing address. Thanks! AAAI 445 Burgess Drive Menlo Park, CA 94025 (415) 328-3123 AAAI-Office@sumex-aim.stanford.edu ------------------------------ Date: 5 Jun 1987 16:45:12-EDT (Friday) From: "Victor S. Miller" Reply-to: THEORYNT%YKTVMX.BITNET@forsythe.stanford.edu Subject: Number Theory Net [Forwarded from the Stanford bboard.] Announcing Number Theory Net I would like to start a separate network for Number Theorists around the world. This would be similar in principle to Theory Net (and probably have some overlap). The purpose of NumberTheory Net would be to help communication among those who work in number theory. Appropriate submissions would be: problems, solutions, queries, notification of address changes, announcement of results, etc. For now, all submissions will be handled by the same userids as for TheoryNet: TheoryNet@ibm.com or theorynt@yktvmx.bitnet for submissions, and TheoryNet-Request@ibm.com or theorynt@yktvmx.bitnet for administrative matters (e.g. additions or deletions to the subscriber list, requests for back submissions, etc.). All contributions should be clearly labeled as being for NumberTheoryNet. I think that it should be useful and enlightening. Victor S. Miller -- moderator ------------------------------ Date: 5 Jun 87 13:52:47 GMT From: salveter@bu-cs.bu.edu (Sharon Salveter) Subject: Need small expert system for research I am directing a research project on knowledge transfer for expert systems. Essentially, we are trying to automate the function of the knowledge engineer in classification-type expert systems. We are looking for small (fewer than 1000 rules) existing expert systems to use as our domains and testbeds/benchmarks. If you have such a system that you would like to donate to research, please contact me. Sharon Salveter Computer Science Boston University ------------------------------ Date: 2 Jun 87 22:10:53 GMT From: imagen!auspyr!dlb!dana!rap@ucbvax.Berkeley.EDU (Rob Peck) Subject: Speech Data Compression I am interested in finding some kind of data compression algorithm that is suitable for compressing speech data. As I understand it, human speech has a great deal of redundancy to it, i.e. repetitions of virtually the same waveforms over a period of time, as well as slow changes in many cases from one waveform to the next. However, if one takes a set of audio samples of a spoken word, the samples will not fall in the right spots to show up any such redundancy. Thus, for a simplistic compression algorithm that looks for repeated sequences, no opportunity to compress would be noticed. Could someone point me to the appropriate literature? Or is there some public domain source code that is already available for this? The code needn't be fast on the analysis and compression. On playback, it should be pretty easy to expand, though. That is, play so many repetitions of this waveform at this sampling rate, then do this next one (or better still, adjust the current waveform until it looks like this new one, as a slewing to the new output... that'd be neat). I've read a little about FFT's, but once calculated, I have no idea how to use it or if it gives me remotely what I am looking for here. Please EMAIL directly to me. I will summarize any interesting responses to the Net. Thanks very much. Rob Peck ...ihnp4!hplabs!dana!rap ------------------------------ Date: 8 Jun 87 19:46:16 GMT From: ai!gautier@rsch.wisc.edu (Jorge Gautier) Subject: WANTED: references on the game of dominoes I am looking for references on computer implementations of the game of dominoes. I suspect there are many variations on the rules for this game, but any pointers to papers, commercial products, Ph.D. theses :-), etc. would be much appreciated. Please reply by mail. Jorge Gautier gautier@ai.wisc.edu "America is waiting for a message of some sort or another." ------------------------------ Date: Tue, 09 Jun 87 11:26:42 n From: DAVIS%EMBL.BITNET@wiscvm.wisc.edu Subject: digging up the garbage.... Ok, here's a quick query for old timers on the AIList (no prizes KIL for being the first in line). Did the list ever show much of a response to Doug Hofstadter's "Waking up From the Boolean Dream or, Subcognition as Computation" ? Yes, yes - I know thats it a wet, cloudy, amorphous piece of writing, utterly unpublishable in any journal beginning with the name "Transactions of....", but nevertheless, Hoftstadter's criticism of 'traditional' AI ("high church computationalism") still seems well in place amidst the countless "has anyone seen expert system EXSYS yet ?" and "any clues on dealing with uncertainty within the context of a WHITEWASH based frame-solving fourth generation language ?" that dominate the list...... I don't want to dig up the past, but if it hasn't happened before, are there any defenders of the EXSYS/4GL/"fuzzy reasoning"/etc., etc., approach willing to correct my impressions of the right direction for movement ? Or even just give me a few recent, decent rebuffs to Hofstadter's viewpoint ? yours in statistical emergence, paul davis "i wash my own clothes, i wash my own hair, i wash my own opinions" nb: but my employers provide the washing machine, the shower & the computer. davis@embl.bitnet ------------------------------ Date: 6 Jun 87 01:38:26 GMT From: mnetor!yetti!unicus!craig@seismo.css.gov (Craig D. Hubley) Subject: Taking AI models and applying them to biology... Forgive the wide cross posting, net.gods, but I am interested in gathering an opinion from biological and artifical intelligence people on a model that arises from AI but has (possibly) biological implications: Foreword or WHY I'M WRITING THIS. -------------------------------- I was semi-surprised in recent months to discover that cognitive psychology, far from developing a bold new metaphor for human thinking, has (to a degree) copied at least one metaphor from third-generation computer science. This description of the human memory system, though cloaked in vaguer terms, corresponds more or less one-to-one with the traditional computer architecture we all know and love. To wit: - senses have "iconic" and "echo" memories analogous to buffers. - short term memory holds information that is organized for quick processing, much like main storage in a computing system. - long term memory holds information in a sort of semantic association network where large related pieces of information reside, similar to backing or "archived" computing storage. At least this far, this theory appears to owe a lot to computer science. Granted, there is lots of empirical evidence in favour, but we all know how a little evidence can go far too far towards developing an analogy. What I think we may need are good parallel connectionist computing models for the social sciences to copy, rather than these old ones that we are beginning to fuse and modify and discard. After all, engineering can construct and test artifacts much quicker than psychologists can. And investigate their insides and their performance as well... The Point or WHAT I'M THINKING ABOUT ------------------------------------ Single cells are constructed according to instructions resident in their own DNA. When their reproductive process fails, they die, become cancerous, etc... In computing terms, a self-reproducing program messes up the code and therefore fails to function (it does not reproduce). Or, it may continue to reproduce a flawed cell (cancer...). But a biological mechanism such as, say, a muscle or a brain is a massively parallel system consisting of many many redundant cells, each of which is capable of performing (at least almost) the same function. So many many parts would have to fail before the effect was enough to endanger the system as a whole. That is, it degrades gracefully. This effect has been observed in parallel sensing systems, which use several low-resolution phased fields that redundantly cover the same area. Removing one such field results in a loss of resolution, but not utter failure to detect a stimulus. Details in Geoffrey Hinton and others... (Byte AI issue, 1985?) At some point of degradation, the whole parallel system will collapse. Or an aged human being will die of a cold. The Question or WHAT DO YOU THINK? ---------------------------------- Apparently, all human organ weights begin to decline shortly after puberty. The cumulative effect of this seeming reduction of resources isn't felt so strongly until middle-age, when we become more susceptible to disease. So far, this is just a statement of the nature of parallel systems. But does it hold up as a theory of aging? - Is mitosis sufficiently prone to failure to account for organ decline? - Statistically, one would expect exponential distribution for failure of single cells, the rate dependent on mitosis failure, and perhaps modified by other cell-killing factors - Does organ failure, medically, occur at the point where a parallel processing system, mathematically, would fail? I've heard that mammal cells appear to suffer a "hard" reproductive limit of 52 mitosis operations, and that meiosis "resets this counter" to 0. - any comment on this, bio-med types? Is it true? - Would a theory assuming a simple variable or random "counter" in each cell limiting its reproductive span better explain aging (programmed cells...) It doesn't seem so... regardless of the origin of the failure, the observed degradation of the system as a whole would still follow this pattern. The upshot of this is that a potentially useful life science model may have just materialized in artificial intelligence. The main flaw that I can see in it is that a cell is complex mechanism in and of itself, and so the success/failure of each might be subject to many factors in parallel as well. That is, it might not fail the way a short subroutine would were it copied badly, which is the gist of this. But then one might find a lower level where the parts were sufficiently monolithic that the analogy held. This seems to kick the butt of the good old 'Entropy' theory... cop-out. Incompentent nineteenth century philosophers leaned heavily on entropy. Comments? Flames? The name of a good shrink? Musing, Craig. ------------------------------ Date: 10 Jun 87 02:49:55 GMT From: hao!boulder!pell@husc6.harvard.edu (Anthony Pelletier) Subject: Re: Taking AI models and applying them to biology... (Craig D. Hubley) writes: (cognative psycology) >far from developing a bold new metaphor for human thinking, has (to a degree) >copied at least one metaphor from third-generation computer science. > one of the things that has always amused me is that, to the extent that I understand the structuring of computers, it seems that the cell and the computer scientists have come up with similar solutions to many of the same questions. This is particularly true when one looks at information flow in the cell. I feel comfortable in assuming that the cell had little help from the CS types in solving problems of information flow. It is likely to be true that contemporaries of in different scientific fields play with each other's ideas. This is why "Nature" insists on being so broad and why F.H.C.C. can get work. But I should stay more to the point. >The Question or WHAT DO YOU THINK? >---------------------------------- >Apparently, all human organ weights begin to decline shortly after puberty. >The cumulative effect of this seeming reduction of resources isn't felt so >strongly until middle-age, when we become more susceptible to disease. >- Is mitosis sufficiently prone to failure to account for organ decline? > >I've heard that mammal cells appear to suffer a "hard" reproductive limit >of 52 mitosis operations, and that meiosis "resets this counter" to 0. > It would seem to me that the step that is likely to give the cell trouble is not mitosis but DNA replication. If a whole chromosome lost or non-disjoined, that cell is in some serious trouble. Progressive accumulation mistakes through replication and general maintanence seems a more likely culprit. I confess that once the topic turns to outside the single cell or involves more than, say, two cells, I am hopelssly lost. So the question of aging is outside my capabilities. This will not, of course, stop me from volenteering the following: I have never liked the "hard-wired-number-of-mitosis" model. I am not sure why; it just seems implausible, or worse yet, unecessary. Supposedly "immortal" cells, like bacteria, actually have a rather high death rate in the population (try doing a particle count then plating them out to see how many are actually able to continue dividing). Their apparent immortality is the result of unrestrained growth. I suspect the failure rate is similar between bacteria and individual cells of a metazoan. The difference may be simply that a metazoan cannot tolerate unrestrained growth of cell populations. The cells are forced to stop dividing when in contact with other cells. they can be induced to re-enter the cycle by growth factors released, for example, when the skin is cut. I would guess that if one coupled the limitations on growth necessary to be a metazoan with accumulated errors, both during replication and simple maitanence, one could explain gradual breakdown of tissue without invoking the "hard-wire" model. oh well, I've gone on too long already. tony (few degrees are worth remembering--and none are worth predicting) Pelletier Molecular etc. Bio Boulder, Co. 80309-0347 P.S. I think alot about information flow problems and would enjoy discussions on that...if anyone wants to chat. ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jun 11 03:21:50 1987 Date: Thu, 11 Jun 87 03:21:39 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #140 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 11 Jun 87 03:10 EDT Received: from relay.cs.net by RELAY.CS.NET id ah07121; 10 Jun 87 12:12 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa09561; 10 Jun 87 12:10 EDT Date: Wed 10 Jun 1987 00:22-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #140 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Wednesday, 10 Jun 1987 Volume 5 : Issue 140 Today's Topics: AI Tools - ID3 vs C4 & Expert Systems for CAD, Comment - Precision in Writing, Theory - Complexity Theory & Applying AI Models to Biology & Symbol Grounding and Physical Invertibility ---------------------------------------------------------------------- Date: 4 Jun 87 13:04:17 GMT From: reiter@endor.harvard.edu (Ehud Reiter) Subject: Re: ID3 vs C4 In article <114@upba.UUCP> damon@upba.UUCP (Damon Scaggs) writes: >I understand that Ross Quinlan, author of the ID3 classification algorithm >has developed a better version with the designation C4. I am looking for >any papers or references about this new algorithm as well as any comments >about what it does better. The best reference I've seen on statistical algorithms for learning decision trees is CLASSIFICATION AND REGRESSION TREES by L. Breiman, J. Friedman, R. Olshen, C. Stone Wadsworth Press, 1984 The book makes no specific mention of ID3 or C4, but it gives much more detail about this class of learning algorithms than I've seen in any of Quinlan's papers. I'm posting this reponse to the net because I really do think this is a superb book. Ehud Reiter reiter@harvard (ARPA,BITNET,UUCP) reiter@harvard.harvard.EDU (new ARPA) ------------------------------ Date: 4 Jun 87 19:16:24 GMT From: bolasov@athena.mit.edu (Benjamin I Olasov) Reply-to: aphrodite!bolasov@RUTGERS.EDU (Benjamin I Olasov) Subject: Re: Wanted: Information on current work in Expert Systems for CAD In article <8705300459.AA08391@ucbvax.Berkeley.EDU> SPANGLER@gmr.COM writes: >I am beginning a survey of the current status of work in applying Expert >Systems technology to Computer Aided Design. This survey is being done >for the Knowledge Engineering group at General Motors. > >I would greatly appreciate any descriptions of or references to research >in this area, as well as information on what CAD expert systems and >expert system shells are available for purchase. > > -- Scott Spangler, spangler@gmr.com > -- Advanced Engineering Staff, GM I just finished writing a master's thesis on just this topic, however it focuses primarily on applications for architectural practice, especially design with a structural pre-cast concrete panel system. You may also write or send e-mail to Professor Sriram in the Civil Engineering Department here at MIT who has written an expert system for structural design called DESTINY. His e-mail address is sriram@athena.mit.edu- mine is bolasov@aphrodite.mit.edu or Olasov@MIT-MULTICS.ARPA. Good luck! Ben ------------------------------ Date: Tue, 09 Jun 87 11:26:05 n From: DAVIS%EMBL.BITNET@wiscvm.wisc.edu Subject: precision in writing I have no wish to make a big deal out of this point, but I feel that Laura Creighton's remarks on precision in writing/expression must be dealt with. She writes: > Precision is a good thing. If one can say precisely what one means then > one will not be misunderstood. This, alas, is a pipe dream. There > is no way to say precisely what one means -- what one says does not > have precise meaning embedded in the words or the relationships between > the words. Rather, one shares a linguistic context with one's > audience. This means that the serach for precision is never ending. I'm afraid that there is a gross difference between the precise delineation of an idea, and over-precise word usage. To be sure, all of human activity is constantly capable of generating new words, and new uses for old words (radical! barf! hack! bug!) - but this alone does not justify the `jargonising' of debate. I believe that if two (or more) people wish to debate any issue, then they have a responsibility to do so on as much common ground as humanly possible. You think that Bertrand Russel was any less capable of a meaningful debate on various aspects of philosphy/cognition because he didn't have access to computerese ? The delineation of an idea is capable of being precise through carefully chosen analogy and metaphor. Such a route is actually better than jargonising since the writer/speaker stands a better chance of getting the audience to appreciate the *core* of an idea, rather than sit back satisfied that they *think* they understand his words...... Sorry to go on on this one, but so much of the debate in and around AI/cognitive science/philosphy of mind gets bogged down by people jargonising their positions, which forces replies to first hack through the cloud that surrounds potentially good (or bad) opinions. yours for jargon free AI, paul davis "i wash my own clothes, i wash my own hair, i wash my own opinions" nb: but my employers provide the washing machine, the shower & the computer davis@embl.bitnet ------------------------------ Date: 3 Jun 87 17:15:21 GMT From: mcvax!botter!klipper!biep@seismo.css.gov (J. A. "Biep" Durieux) Subject: What philosophical problems does complexity theory yield? Suppose P != NP. Then some things will take a long time to compute. But so what? Suppose someone finds out not all problems can be solved in constant time. Now that comes as a philosophical shock, of course. That has lots of implications. But once one has overcome that shock, finding that some problems cannot be solved in linear time may be annoying, but now since the possibility of constant time already has been destroyed, it's no great news. As, one by one, all sorts of upper bounds on exponents prove false, and finally it seems polynomial time isn't good at all, one gets even bored by all those variations on the same theme, not? So what exactly is so exciting about that polynomial limit? About constant time solutions: Seemingly linear-time solutions can often be turned into constant-time solutions by applying parallelism. This is the way the universe is able to simulate itself, however big it (object) may be or grow. I don't know of what complexity the collapsing of a wave- function would be supposed that all "time-space-points" of it worked parallelly on it. But, isn't anything which cannot be turned into a constant-time process philosophically annoying? Why just hassling about non-polynomial time solutions? Am I missing something? (Shouldn't I have asked that at the beginning of this article? :-)) -- Biep. (biep@cs.vu.nl via mcvax) Popper tells us how science ought to be done - Kuhn tells us how it *is* done. And I ... I read neither. ------------------------------ Date: 4 Jun 87 16:09:44 GMT From: ramesh@cvl.umd.edu (Ramesh Sitaraman) Subject: Re: What philosophical problems does complexity theory yield? In article <789@klipper.cs.vu.nl> biep@cs.vu.nl (J. A. "Biep" Durieux) writes: >But, isn't anything which cannot be turned into a constant-time process >philosophically annoying? Why just hassling about non-polynomial time >solutions? Am I missing something? (Shouldn't I have asked that at the >beginning of this article? :-)) >-- Yes, you are missing the point !! The difference between a polynomial and non-polynomial solution for a problem is the difference between structure and a complete lack of it. If P not = NP we would have shown that some problems can be solved only by something similar to a dumb exhaustive search over the solution space i.e. there is not enough structure in the problem to constrain its solutions. Graph theorists have found eulerian circuits very interesting and there have been very strong theorems proved about graphs with this property. However, the seemingly similar problem of Hamiltonian circuits have almost no characterisation inspite of dilligent efforts for the past 100 years or so. The theory of NP-completeness explains this anomaly. Eulerian circuit can be solved in linear time while Hamiltonian circuit is NP-complete !!! Ramesh (Defn: Eulerian ckt: A circuit passing through all edges of a graph without repeating an edge. Hamiltonian ckt: A circuit passing through all the vertices of a graph without repeating a vertex. ) ------------------------------ Date: 6 Jun 87 20:52:53 GMT From: umnd-cs!umn-cs!moll@ucbvax.Berkeley.EDU (Rick Moll) Subject: Re: What philosophical problems does complexity theory yield? In article <789@klipper.cs.vu.nl> biep@cs.vu.nl (J. A. "Biep" Durieux) writes: >Suppose P != NP. Then some things will take a long time to compute. >But so what? >As, one by one, all sorts of upper bounds on exponents prove false, and[...] You seem to be implying that if P=NP then *all* problems can be solved in polynomial time. This is certainly not so. Given any computable function f(x), one can construct (by diagonalization) a problem which can be solved, but cannot be solved in time f(n) on a Turing machine. I believe the same proof would work for parallel machines. >About constant time solutions: Seemingly linear-time solutions can often be >turned into constant-time solutions by applying parallelism. [...] Note that P=NP is stated as a problem about Turing machines (or sometimes single processor random access machines). Any problem in the class NP can definately be solved in polynomial time is one is allowed to use an arbitrary number (varying with the size of the problem instance) of processors. ------------------------------ Date: 7 Jun 87 00:28:36 GMT From: chiaraviglio@husc4.harvard.edu (lucius chiaraviglio) Subject: Re: Taking AI models and applying them to biology... In article <622@unicus.UUCP> craig@unicus.UUCP (Craig D. Hubley) writes: >I've heard that mammal cells appear to suffer a "hard" reproductive limit >of 52 mitosis operations, and that meiosis "resets this counter" to 0. > >- any comment on this, bio-med types? Is it true? >- Would a theory assuming a simple variable or random "counter" in each cell >limiting its reproductive span better explain aging (programmed cells...) Random failure may be a significant factor in aging, but a hard limit on the number of times a cell may divide before it self-destructs has been observed in tissue culture, where the cells are for the most part not dependant on each other. Those cells which manage to get past the hard limit are abnormal (although not necessarily cancerous) in ways beyond their mere ability to keep dividing after they were supposed to self-destruct. I don't remember most of the details of this, but I do remember that they tend to become tetraploid (I think also aneuploid) due to an increase in the rate of mitotic failure. -- Lucius Chiaraviglio lucius%tardis@harvard.harvard.edu seismo!tardis.harvard.edu!lucius ------------------------------ Date: 9 Jun 87 00:02:19 GMT From: pixar!davel@ucbvax.Berkeley.EDU (David Longerbeam) Subject: Re: Taking AI models and applying them to biology... In article <622@unicus.UUCP>, craig@unicus.UUCP (Craig D. Hubley) writes: > > This description of the human memory system, though cloaked in vaguer terms, > corresponds more or less one-to-one with the traditional computer > architecture we all know and love. To wit: [description deleted] > At least this far, this theory appears to owe a lot to computer science. > Granted, there is lots of empirical evidence in favour, but we all know > how a little evidence can go far too far towards developing an analogy. One of my philosophy professors in college offered the observation that models for the human mind have always seemed to correspond to the most advanced form of technology at that given point in history. He could recall that when he was young, this technology was the combustion engine, and lo, the cognitive psychologists' model at that time was the combustion engine. Of course, this technology is now the digital computer, and many psychologists, linguists and computer scientists use it as a model to explain activites of the human mind. Some go so far as to say that intelligence is nothing more than the result of following the same sorts of syntactical rules as performed by a computer! But I stray... I wanted to point out that you didn't give the source of the above model/ comparison, and that if it is not entirely empirical in nature, it may be a case of "use the latest technology as the best model". -- David Longerbeam || The opinions expressed above Pixar || are not to be contrued as the San Rafael, CA || opinions, stated or otherwise, ucbvax!pixar!davel (415) 499-3600 || of Pixar. ------------------------------ Date: 5 Jun 87 04:45:45 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: symbol grounding and physical invertibility John Cugini asks: > (1) I wonder why the grounding is to depend on invertibility rather than > causation and/or resemblance? > > (2) Isn't it true that physically distinct > kinds of light (eg. #1 red-wavelength and green-wavelength vs. > #2 yellow-wavelength) can cause completely indistinguishable > sensations (ie subjective yellow)? Is this not, then, a non-invertible, > but nonetheless grounded sensation? (1) According to my view, invertibility (and perhaps inversion) captures just the relevant features of causation and resemblance that are needed to ground symbols. The relation is between the proximal projection (of a distal object) onto the sensory surfaces -- let's call it P -- and an invertible transformation of that projection [I(P)]. The latter is what I call the "iconic representation." Note that the invertibility is with the sensory projection, *not* the distal object. I don't believe in distal magic. My grounding scheme begins at the sensory surfaces ("skin and in"). No "wider" metaphysical causality is involved, just narrow, local causality. Of course the story is more complicated, because iconic representations are not sufficient to ground a symbol referring to an object. They're not even enough to allow a device to reliably pick out the object and give it the right name (i.e., to categorize or identify it). "Categorical representations" are needed next, but these are no longer invertible into the sensory projection. They are feature-filters preserving only the (proximal) properties of the object's sensory projection that reliably distinguish the object (let's say it's an "X") from the other objects that it can be confused with (i.e., relevant "non-X's" in the particular context of confusable alternatives sampled to date). Then finally the labels ("X," "non-X") can be used as the primitive symbols in a (now *grounded*) symbol system, to be combined and otherwise syntactically manipulated into meaningful composite symbol-strings (descriptions). (2) Your question about indistinguishable but distinct colors mistakes my grounding scheme for a "wide" metaphysical grounding scheme -- one where the critical "causality" would be in the relation between distal objects and our internal representations of them, whereas mine is a narrow, skin-and-in grounding proposal. I have dubbed this view "approximationism," and, without going into details (for which you may want to consult the CP book or a reprint of the theoretical chapter), the essence of the idea is that internal representations are always approximate rather than "exact," in two important senses. The iconic representation is approximate up to its grain of resolution (the "jnd" or "just-noticeable-difference"): Think of it as a Principle of the "Iconic Identity of Iconic Indiscernibles": What you can't tell apart is the same to you. The categorical representations are approximate in an even more important sense: The only features the category filter picks out are the ones that are needed in order to identify the confusable alternatives in the context you have sampled to date. Hence an X is just what your current, approximate, provisional context-dependent feature-filter picks out reliably from among the X's and Non-X's you have encountered so far: "The Categorical Identity of Unsorted or Unsortable Members" (i.e., X's are identically X's unless and until reliably identified or identifiable otherwise). Since this is not a "wide" grounding, there is nothing oracular or omniscient or metaphysical about what the X is that this picks out. There is no God's-eye view from which you can say what X's "really" are. There's just the mundane historical fact -- available to an outside observer, if there is one -- about what the actual distal objects were whose proximal projections you were sampling. Those furnished your context, and your fallible, context-dependent representations will always be approximate relative to those objects. In conclusion, the only differences in the object that are reflected in the iconic and categorical representations are the ones present in the proximal projection of the alternatives sampled to date (and preserved by the category-feature filter). The representations are approximate (i.e., indifferent) with respect to any further distal differences. Symbolic discourse may serve to further tighten the approximation, but even that cannot be "exact," if for no other reason than that there's always a tomorrow, in which the context may be widened and the current representation may have to be revised. -- But that's another story, and no longer concerns the grounding problem but what's called "inductive risk." -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jun 11 03:21:30 1987 Date: Thu, 11 Jun 87 03:21:22 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #141 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 11 Jun 87 03:07 EDT Received: from relay.cs.net by RELAY.CS.NET id ac04630; 10 Jun 87 3:59 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa06138; 10 Jun 87 3:57 EDT Date: Wed 10 Jun 1987 00:29-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #141 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Wednesday, 10 Jun 1987 Volume 5 : Issue 141 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 5 Jun 87 17:12:10 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In reply to my objection that >> invertibility has essentially *nothing* to do with the difference >> between analog and digital representation according to anybody's >> intuitive use of the terms Stevan Harnad (harnad@mind.UUCP) writes in message <792@mind.UUCP>: >There are two stages of A/D even in the technical sense. ... Unless the >original signal is already discrete, the quantization phase involves a >loss of information. Some regions of input variation will not be retrievable >from the quantized image. The transformation ... cannot be inverted so as to >recover the entire original signal. Well, what I think is interesting is not preserving the signal itself but rather the *information* that the signal carries. In this sense, an analog signal conveys only a finite amount of information and it can in fact be converted to digital form and back to analog *without* any loss. But in any case the point I've been emphasizing remains: the A/A transformations you envisage are not going to be perfect (no "skyhooks" now, remember?), so preservation or loss of information alone won't distinguish an (intuitively) A/A from an A/D transfomation. I think the following reply to this point only muddies the waters: > I agree that there may be information loss in >A/A transformations (e.g., smoothing, blurring or loss of some >dimensions of variation), but then the image is simply *not analog in >the properties that have been lost*! It is only an analog of what it >preserves, not what it fails to preserve. You can take this line if you like, but notice that the same is true of a *digitized* image -- in your terms, it is "analog" in the information it preserves and not in the information lost. This seems to me to be a very unhappy choice of terminology! Both analog and digitizing transformations must preserve *some* information. If all you're *really* interested in is the quality of being (naturally) information-preserving (i.e. physically invertible), than I'd strongly recommend you just use one of these terms and drop the misleading use of "analog", "iconic", and "digital". > The "symbol grounding problem" that has >been under discussion here concerns the fact that symbol systems >depend for their "meanings" on only one of two possibilities: One is >an interpretation supplied by human users... and the other is a physical, >causal connection with the objects to which the symbols refer. >The surprising consequence is that a "dedicated system" -- one that is >hard-wired to its transducers and effectors... may be significantly different >from the very *same* system as an isolated symbol-manipulating module, >cut off from its peripherals ... With regard to this "symbol grounding problem": I think it's been well-understood for some time that causal interaction with the world is a necessary requirement for artificial intelligence. Recall that in his BBS reply to Searle, Dennett dismissed Searle's initial target -- the "bedridden" form of the Turing test -- as a strawman for precisely this reason. (Searle believes his argument goes through for causally embedded AI programs as well, but that's another topic.) The philosophical rationale for this requirement is the fact that some causal "grounding" is needed in order to determine a semantic interpretation. A classic example is due to Georges Rey: it's possible that a program for playing chess could, when compiled, be *identical* to one used to plot strategy in the Six Day War. If you look only at the formal symbol manipulations, you can't distinguish between the two interpretations; it's only by virtue of the causal relations between the symbols and the world that the symbols could have one meaning rather than another. But although everyone agrees that *some* kind of causal grounding is necessary for intentionality, it's notoriously difficult to explain exactly what sort it must be. And although the information-preserving transformations you discuss may play some role here, I really don't see how this challenges the premises of symbolic AI in the way you seem to think it does. In particular you say that: >The potential relevance of the physical invertibility criterion >would only be to cognitive modeling, especially in the constraint that >a grounded symbol system must be *nonmodular* -- i.e., it must be hybrid >symbolic/nonsymbolic. But why must the arrangement you envision must be "nonmodular" ? A system may contain analog and digital subsystems and still be modular if the subsytems interact solely via well-defined inputs and outputs. More importantly -- and this is the real motivation for my terminological objections -- it isn't clear why *any* (intuitively) analog processing need take place at all. I presume the stance of symbolic AI is that sensory input affects the system via an isolable module which converts incoming stimuli into symbolic representations. Imagine a vision sub-system that converts incoming light into digital form at the first stage, as it strikes a grid of photo-receptor surfaces, and is entirely digital from there on in. Such a system is still "grounded" in information-preserving representations in the sense you require. In short, I don't see any *philosophical* reason why symbol-grounding requires analog processing or a non-modular structure. Anders Weinstein BBN Labs ------------------------------ Date: 7 Jun 87 18:25:00 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., Cambridge, MA writes: > [regarding invertibility, information preservation and the A/D > distinction]: what I think is interesting is not preserving the > signal itself but rather the *information* that the signal carries. > In this sense, an analog signal conveys only a finite amount of > information and it can in fact be converted to digital form and back > to analog *without* any loss. This is an important point and concerns a matter that is at the heart of the symbolic/nonsymbolic issue: What you're saying is appropriate for ordinary communication theory and communication-theoretic applications such as radio signals, telegraph, radar CDs, etc. In all these cases the signal is simply a carrier that encodes information which is subsequently decoded at the receiving end. But in the case of human cognition this communication-theoretic model -- of signals carrying messages that are encoded/decoded on either end -- may not be appropriate. (Formal information theory has always had difficulties with "content" or "meaning." This has often been pointed out, and I take this to be symptomatic of the fact that it's missing something as a candidate model for cognitive "information processing.") Note that the communication-theoretic, signal-analytic view has a kind of built-in bias toward digital coding, since it's the "message" and not the "medium" that matters. But what if -- in cognition -- the medium *is* the message? This may well be the case in iconic processing (and the performances that it subserves, such as discrimination, similarity judgment, matching, short-term memory, mental rotation, etc.): It may be the structure or "shape" of the physical signal (the stimulus) itself that matters, not some secondary information or message it carries in coded form. Hence the processing may have to be structure- or shape-preserving in the physical analog sense I've tried to capture with the criterion of invertibiliy. > a *digitized* image -- in your terms... is "analog" in the > information it preserves and not in the information lost. This > seems to me to be a very unhappy choice of terminology! Both analog > and digitizing transformations must preserve *some* information. > If all you're *really* interested in is the quality of being > (naturally) information-preserving (i.e. physically invertible), > than I'd strongly recommend you just use one of these terms and drop > the misleading use of "analog", "iconic", and "digital". I'm not at all convinced yet that the sense of iconic and analog that I am referring to is unrelated to the signal-analytic A/D distinction, although I've noted that it may turn out, on sufficient analysis, to be an independent distinction. For the time being, I've acknowledged that my invertibility criterion is, if not necessarily unhappy, somewhat surprising in its implications, for it implies (1) that being analog may be a matter of degree (i.e., degree of invertibility) and (2) even a classical digital system must be regarded as analog to a degree if one is considering a larger "dedicated" system of which it is a hard-wired (i.e., causally connected) component rather than an independent (human-interpretation-mediated) module. Let me repeat, though, that it could turn out that, despite some suggestive similarities, these considerations are not pertinent to the A/D distinction but, say, to the symbolic/nonsymbolic distinction -- and even that only in the special context of cognitive modeling rather than signal analysis or artificial intelligence in general. > With regard to [the] "symbol grounding problem": I think it's been > well-understood for some time that causal interaction with the world > is a necessary requirement for artificial intelligence... > The philosophical rationale for this requirement is the fact that > some causal "grounding" is needed in order to determine a semantic > interpretation... But although everyone agrees that *some* kind of > causal grounding is necessary for intentionality, it's notoriously > difficult to explain exactly what sort it must be. And although the > information-preserving transformations you discuss may play some role > here, I really don't see how this challenges the premises of symbolic > AI in the way you seem to think it does. As far as I know, there have so far been only two candidate proposals to overcome the symbol grounding problem WITHOUT resorting to the kind of hybrid proposal I advocate (i.e., without giving up purely symbolic top-down modules): One proposal, as you note, is that a pure symbol-manipulating system can be "grounded" by merely hooking it up causally in the "right way" to the outside world with simple (modular) transducers and effectors. I have conjectured that this strategy will not work in cognitive modeling (and I have given my supporting arguments elsewhere: "Minds, Machines and Searle"). The strategy may work in AI and conventional robotics and vision, but that is because these fields *do not have a grounding problem*! They're only trying to generate intelligent *pieces* of performance, not to model the mind in *all* its performance capacity. Only cognitive modeling has a symbol grounding problem. The second nonhybrid way to try to ground a purely symbolic system in real-world objects is by cryptology. Human beings, knowing already at least one grounded language and its relation to the world, can infer the meanings of a second one [e.g., ancient cuneiform] by using its internal formal structure plus what they already know: Since the symbol permutations and combinations of the unknown system (i.e., its syntactic rules) are constrained to yield a semantically interpretatable system, sometimes the semantics can be reliably and uniquely decoded this way (despite Quine's claims about the indeterminacy of radical translation). It is obvious, however, that such a "grounding" would be derivative, and would depend entirely on the groundedness of the original grounded symbol system. (This is equivalent to Searle's "intrinsic" vs. "derived intentionality.") And *that* grounding problem remains to be solved in an autonomous cognitive model. My own hybrid approach is simply to bite the bullet and give up on the hope of an autonomous symbolic level, the hope on which AI and symbolic functionalism had relied in their attempt to capture mental function. Although you can get a lot of clever performance by building in purely symbolic "knowledge," and although it had seemed so promising that symbol-strings could be interpreted as thoughts, beliefs, and mental propositions, I have argued that a mere extension of this modular "top-down" approach, hooking up eventually with peripheral modules, simply won't succeed in the long run (i.e., as we attempt to approach an asymptote of total human performance capacity, or what I've called the "Total Turing Test") because of the grounding problem and the nonviability of the two "solutions" sketched above (i.e., simple peripheral hook-ups and/or mediating human cryptology). Instead, I have described a nonmodular hybrid representational system in which symbolic representations are grounded bottom-up in nonsymbolic ones (iconic and categorical). Although there is a symbolic level in such a system, it is not quite the autonomous all-purpose level of symbolic AI. It trades its autonomy for its groundedness. > [W]hy must the arrangement you envision be "nonmodular"? A system > may contain analog and digital subsystems and still be modular if > the subsystems interact solely via well-defined inputs and outputs. I'll try to explain why I believe that a successful mind-model (one able to pass the Total Turing Test) is unlikely to consist merely of a pure symbol-manipulative module connected to input/output modules. A pure top-down symbol system just consists of physically implemented symbol manipulations. You yourself describe a typical example of ungroundedness (from Georges Rey): > it's possible that a program for playing chess could, > when compiled, be *identical* to one used to plot > strategy in the Six Day War. If you look only at the > formal symbol manipulations, you can't distinguish between > the two interpretations; it's only by virtue of the causal > relations between the symbols and the world that the symbols > could have one meaning rather than another. Now consider two cases of "fixing" the symbol interpretations by grounding the causal relations between the symbols and the world. In (1) a "toy" case -- a circumscribed little chunk of performance such as chess-playing or war-games -- the right causal connections could be wired according to the human encryption/decryption scheme: Inputs and outputs could be wired into their appropriate symbolic descriptions. There is no problem here, because the toy problems are themselves modular, and we know all the ins and outs. But none but the most diehard symbolic functionalist would want to argue that such a simple toy model was "thinking," or even doing anything remotely like what we do when we accomplish the same performance. The reason is that we are capable of doing *so much more* -- and not by an assemblage of endless independent modules of essentially the same sort as these toy models, but by some sort of (2) integrated internal system. Could that "total" system be just an oversized toy model -- a symbol system with its interpretations "fixed" by a means analogous to these toy cases? I am conjecturing that it is not. Toy models don't think. Their internal symbols really *are* meaningless, and hence setting them in the service of generating a toy performance just involves hard-wiring our intended interpretations of its symbols into a suitable dedicated system. Total (human-capacity-sized) models, on the other hand, will, one hopes, think, and hence the intended interpretations of their symbols will have to be intrinsic in some deeper way than the analogy with the toy model would suggest, at least so I think. This is my proposed "nonmodular" candidate: Every formal symbol system has both primitive atomic symbols and composite symbol-strings consisting of ruleful combinations of the atoms. Both the atoms and the combinations are semantically interpretable, but from the standpoint of the formal syntactic rules governing the symbol manipulations, the atoms could just as well have been undefined or meaningless. I hypothesize that the primitive symbols of a nonmodular cognitive symbol system are actually the (arbitrary) labels of object categories, and that these labels are reliably assigned to their referents by a nonsymbolic representational system consisting of (i) iconic (invertible, one-to-one) transformations of the sensory surface and (ii) categorical (many-to-few) representations that preserve only the features that suffice to reliably categorize and label sensory projections of the objects in question. Hence, rather than being primitive and undefined, and hence independent of interpretation, I suggest that the atoms of cognitive symbol systems are grounded, bottom-up, in such a categorization mechanism. The higher-order symbol combinations inherit the bottom-up constraints, including the nonsymbolic representations to which they are attached, rather than being an independent top-down symbol-manipulative module with its connections to an input/output module open to being fixed in various extrinsically determined ways. > it isn't clear why *any* (intuitively) analog processing need > take place at all. I presume the stance of symbolic AI is that > sensory input affects the system via an isolable module which converts > incoming stimuli into symbolic representations. Imagine a vision > sub-system that converts incoming light into digital form at the > first stage, as it strikes a grid of photo-receptor surfaces, and is > entirely digital from there on in. Such a system is still "grounded" > in information-preserving representations in the sense you require. > In short, I don't see any *philosophical* reason why symbol-grounding > requires analog processing or a non-modular structure. It is exactly this modular scenario that I am calling into question. It is not clear at all that a cognitive system must conform to it. To get a device to be able to do what we can do we may have to stop thinking in terms of "isolable" input modules that go straight into symbolic representations. That may be enough to "ground" a conventional toy system, but, as I've said, such toy systems don't have a grounding problem in the first place, because nobody really believes they're thinking. To get closer to life-size devices -- devices that can generate *all* of our performance capacity, and hence may indeed be thinking -- we may have to turn to hybrid systems in which the symbolic functions are nonmodularly grounded, bottom-up, in the nonsymbolic ones. The problem is not a philosophical one, it's an empirical one: What looks as if it's likely to work, on the evidence and reasoning available? -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 16 03:21:14 1987 Date: Tue, 16 Jun 87 03:21:07 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #142 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 16 Jun 87 03:06 EDT Received: from relay.cs.net by RELAY.CS.NET id ab07350; 15 Jun 87 2:11 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10375; 15 Jun 87 2:13 EDT Date: Sun 14 Jun 1987 22:45-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #142 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 15 Jun 1987 Volume 5 : Issue 142 Today's Topics: Program - Cognitive Science at Occidental College, Seminars - Universal Plans: Emergent Goal Structures (SRI) & Abductive Reasoning in Multifault Diagnostic Systems (UPenn) & Potential Histories and Inertial Theories (SU) & Controlling Execution of Logic Programs (MCC), Conferences - OOPSLA-87 & Architectures for Intelligent Interfaces ---------------------------------------------------------------------- Date: 04 Jun 87 09:00:45 PST From: oxy!traiger@csvax.caltech.edu (Saul P. Traiger) Subject: Program - Cognitive Science at Occidental College Occidental College, a liberal arts college which enrolls approximately 1600 students, is pleased to announce a new Program in Cognitive Science. The Program offers an undergraduate major and minor in Cognitive Science. Faculty participating in this program include members of the departments of mathematics, linguistics, psychology, and philosophy. The program is the result of one of the most exciting developments in higher education today, namely the interaction among philosophers, mathematicians, psychologists, linguists, and computer scientists. This interaction is the result of common interests in cognitive science. Computer architecture is now as likely to be discussed in a philosophy or psychology seminar as it is in a computer science course. Shared interests in cognitive science lead to the development and adoption of an interdepartmental program in cognitive science at Occidental College. The undergraduate major in Cognitive Science at Occidental College includes courses in mathematics, philosophy, psychology and linguistics. Instruction in mathematics introduces students to computer languages, discrete mathematics, logic, and the mathematics of computation. Philosophy offerings cover the philosophy of mind, with emphasis on computational models of the mind, the theory of knowledge, the philosophy of science, and the philosophy of language. Psychology courses include basic psychology, learning, perception, and cognition. Courses in linguistics provide a theoretical foundation in natural languages, their acquisition, development, and structure. For more information about Occidental College's Cognitive Science Program please contact: Professor Saul Traiger Cognitive Science Program 1600 Campus Road Occidental College Los Angeles, CA 90041 ARPANET: oxy!traiger@CSVAX.Caltech.EDU BITNET: oxy!traiger@hamlet CSNET: oxy!traiger%csvax.caltech.edu@RELAY.CS.NET UUCP: ....{seismo, rutgers, ames}!cit-vax!oxy!traiger ------------------------------ Date: Thu, 11 Jun 87 11:29:04 PDT From: Amy Lansky Subject: Seminar - Universal Plans: Emergent Goal Structures (SRI) EXECUTING UNIVERSAL PLANS: EMERGENT GOAL STRUCTURES & THEIR USES Marcel Schoppers (MARCEL@ADS.COM) Advanced Decision Systems 11:00 AM, MONDAY, June 15 SRI International, Building E, Room EJ228 ``Universal plans'' are designed for execution in unpredictable state spaces, refusing to over-commit to a specific future course of events, and deliberately making no assumptions about how situations might follow one another. Instead, plan synthesis becomes the goal-directed selection of reactions to possible situations; plans become inherently conditional; and plan execution classifies the current situation so as to respond with the selected reaction. Consequently there is no inherent distinction between expected and unexpected events; the concepts of success & failure are irrelevant for both synthesis and execution; and "error recovery" needs no special mechanisms beyond those already present for normal execution. After introducing the Universal Plan representation, this talk will show how at any given instant, plan predicates can be interpreted as goals of achievement or of maintenance, and that this interpretation can be used to reconstruct a four-fold typing of events (of success, failure, serendipity and sabotage). In other words, intentions emerge from the interaction of plan with environment (the environment has a large hand in determining the agent's goals at each moment), and the notions of success and failure are not primitive but perceived (relative to the agent's goals). The Universal Plan representation also indicates precisely which conditions must be monitored at each instant to enable detection of all events of each type. Two benefits follow; I will only mention them briefly. First, we can get complexity estimates for detecting all serendipity and sabotage events, and can produce informed strategies to alleviate sensing costs. Second, the goal structure at each moment in time contains all the information required to choose an appropriate action, thus facilitating incremental synthesis. VISITORS: Please arrive 5 minutes early so that you can be escorted up from the E-building receptionist's desk. Thanks! ------------------------------ Date: Thu, 11 Jun 87 14:13:25 EDT From: tim@linc.cis.upenn.edu (Tim Finin) Subject: Seminar - Abductive Reasoning in Multifault Diagnostic Systems (UPenn) ABDUCTIVE REASONING in MULTIPLE-FAULT DIAGNOSTIC SYSTEMS Gary Morris Computer and Information Science University of Pennsylvania Philadelphia PA Abductive reasoning involves generating explanations for observed facts or symptoms -- i.e. diagnosis. Diagnosis is more difficult, both theoretically and practically, when more than one disorder or fault may occur simultaneously in the system beign diagnosed. Five approaches to this problem are reviewed and contrasted: - Binary Choice Bayesian (Ben-Bassat: the MEDAS system) - Sequential Bayesian (Pople: INTERNIST) - Causal Model Reasoning (Patil: ABEL) - Parsimonious Set Covering (Reggia & Nau: various systems) - "Diagnosis From First Principles" (Reiter, deKleer: various) Finally, an emerging convergence of these methods is described. Friday, June 12, 3:30 pm Room 554 Moore ------------------------------ Date: 01 Jun 87 1605 PDT From: Vladimir Lifschitz Subject: Seminar - Potential Histories and Inertial Theories (SU) [Forwarded from the Stanford bboard.] Yoav Shoham asked me to send a nice little poem to this mailing list: With logics that are monotonic Relations are nice but platonic It's when you permit Just models that fit That things become most erotonic Yoav will also speak at our seminar on a related subject: POTENTIAL HISTORIES AND INERTIAL THEORIES Yoav Shoham Thursday, June 4, 4:15pm Bldg. 160, Room 161K In previous talks I never managed to get to my solution to the extended-prediction problem (which is my name for the problem subsuming the frame problem, a name that, shall we say, never quite caught). I'll describe the intuitive concept of a potential history, which has a strong McDermott-like persistence flavor. I'll then embed the concept formally within the logic of chronological ignorance. I'll identify a class of theories, called inertial theories, which extend causal theories, and yet which a. are expressive enough to capture the notion of potential histories, and b. have the "unique model" and easy computability properties. My intention is this time to go into some detail. I'm still not sure I have enough material for an hour, and if I don't I'll ask the audience some questions on TMS's. ------------------------------ Date: Mon 1 Jun 87 11:04:12-CDT From: Ellie Huck Subject: Seminar - Controlling Execution of Logic Programs (MCC) Madhur Kohli Department of Computer Science University of Maryland June 4 - 10:30am ACA Conference Room 2.806 Controlling the Execution of Logic Programs The performance of a logic programming system is dictated by the control strategy of its problem solving component. This talk describes a methodology for the specification and utilization of control knowledge for logic programs. We describe a control specification system developed as an experimental tool for the study of control issues in problem solving. Analysis of the control behavior of several sequential problem solvers and PRISM, a parallel logic programming system, is used to identify parameters to express control decisions and points at which they apply. These results form the basis for the definition of a control language to specify the control behavior of problem solvers. The language is expressive enough to specify many general and specialized top-down execution schemes for both sequential and parallel problem solvers. A compiler has been developed to generate an interpreter which implements the specified control strategy. Experimental results show that the generated interpreters provide an order of magnitude improvement over meta-interpretation of the control specification. Madhur Kohli June 4 - 10:30 ACA Conference Room 2.806 ------------------------------ Date: 5 Jun 87 18:50:31 GMT From: ut-sally!home.csnet!im4u!ti-csl!fordyce@seismo.CSS.GOV (David Fordyce) Subject: Conference - OOPSLA-87 Article-I.D.: home.444 OBJECT-ORIENTED DATABASE WORKSHOP: Implementation Aspects To be held in conjunction with the Object-oriented Programming Systems, Languages and Applications (OOPSLA-87) Conference October 5, 1987 Orlando, Florida Object-oriented database systems combine the streangths of object-oriented programming systems and data models, and database systems. This half-day workshop will be held on Monday morning, October 5, 1987. The goal of the workshop is to study the implementation aspects of object-oriented database systems. The workshop will focus on issues such as object fault management, storage management (buffering, prefetching, clustering, etc.), object persistence, object sharing, transactions on objects, concurrency control, recovery, and performance issues. The workshop panel will consist of: Timothy Andrewes (Ontologic), Umesh Dayal (Computer Corporation of America), Prof. David Maier (Oregon Graduate Center and Servio Logic), Patrick O'Brien (Digital Equipment Corporation), Prof. Lawrence Rowe (University of California at Berkeley), Prof. Alfred Spector (Carnegie-Mellon University), David Wells (Texas Instruments), and Prof. Stan Zdonik (Brown University). In the first 90 minutes, each panel member will present his position. This will be followed by questions from the workshop participants and discussions. To encourage vigorous interactions and exchange of ideas between the participants, the workshop will be limited to 50 qualified participants. If you are interested in attending the workshop, please submit three copies of a single page abstract to the workshop chairman describing your work related to the implementation issues of object-oriented database systems. The workshop participants will be selected based on the relevance and significance of their work described in the abstract. There will be no proceedings for the workshop. Abstracts should be submitted to the workshop chairman by August 1, 1987. Selected participants will be notified by September 1, 1987 Workshop Chairman: Dr. Satish M. Thatte Manager, Database Systems Branch Artificial Intelligence Laboratory Texas Instruments Incorporated P.O. Box 226015, M/S 238 Dallas, TX 75266 Phone: (214)-995-0340 CSNet: Thatte@TI-CSL -- Regards, David ------------------------------ Date: Fri, 5 Jun 87 09:42:22 PDT From: wiley!sherman@lll-lcc.ARPA (Sherman Tyler) Subject: Conference - Architectures for Intelligent Interfaces Call for Participation Workshop on Architectures for Intelligent Interfaces: Elements and Prototypes March 29 - April 1, 1988, Monterey, California Sponsored by AAAI Objective: The term ``Intelligent Interface'' characterizes the set of computer-human interfaces which employ AI to enhance the transactional nature of the interface. The goal of the workshop is to explore ways in which AI techniques (e.g., knowledge representation, inference mechanisms, and heuristic search) can be used to provide the adaptability and reasoning capabilities required for a more intelligent human-machine interaction. Some possible areas for focused discussions might include: * Models (user, system, task) - adapting the dialogue to the current context of the interaction, considering the particular user, the target system, and the high-level task under execution; * Channels of Communication - allowing users to communicate intentions with a minimum of learning and effort, using Natural Language, Graphics, and the integration of mixed modalities of input; * Planning - for recognizing user plans and their implied goals, generating plans to meet those goals, and planning how to best display the resulting information to communicate the result of the executed action; * Interface-Building Tools - using artificial intelligence techniques to support developers in designing and constructing interfaces. Attendance: In order to provide an intellectually stimulating environment conducive to interaction and exchange of ideas, the attendance will be limited to approximately 35 participants. The ideal participant is an individual who is actively addressing theoretical, research, and/or implementation issues relevant to Intelligent Interfaces (with a bias toward those who have dealt with implementation issues at some level). Limited financial assistance will be available for graduate students who are invited to participate. Review Process: The submitted abstracts and autobiographies will be reviewed by the program committee. Invitation will be based upon relevance of the work to the goals of the workshop, and on the basis of significance, originality, and scientific quality. Workshop Organization: The workshop organizers are J. Sullivan (Lockheed AI Center) and S. Tyler (Lockheed AI Center). The program committee consists of J. Mackinlay (Xerox PARC), R. Neches (USC Information Sciences Institute), E. Rissland (University of Massachusetts), and N. Sondheimer (USC Information Sciences Institute). Submission: A detailed eight page abstract and a one page biographical sketch (six copies of each) should be submitted by September 1, 1987. Invitations for participation will be extended by October 16, 1987, with complete papers due by December 18, 1987. Publication of the proceedings is planned, therefore the quality of the papers is important. Submit abstracts to: Joseph W. Sullivan or Sherman W. Tyler, O/90-06 B/259, Lockheed AI Center, 2710 Sand Hill Rd., Menlo Park, CA 94025, (415) 354-5200, wiley!joe@lll-lcc.arpa or wiley!sherman@lll-lcc.arpa ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Mon Jun 15 03:16:18 1987 Date: Mon, 15 Jun 87 03:16:09 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #143 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Mon, 15 Jun 87 03:10 EDT Received: from relay.cs.net by RELAY.CS.NET id aa07518; 15 Jun 87 2:44 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10475; 15 Jun 87 2:36 EDT Date: Sun 14 Jun 1987 23:18-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #143 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 15 Jun 1987 Volume 5 : Issue 143 Today's Topics: Journal Issues - Neural Networks (IEEE Computer) & Smolensky on Connectionism (BBS) & Laming on Sensory Analysis (BBS), Conference - Genetic Algorithms ---------------------------------------------------------------------- Date: 8 June 1987, 16:25:07 EDT From: Bruce Shriver Subject: Journal Issue - Neural Networks Call for Papers and Referees Special Issue of Computer Magazine on Neural Networks The March, 1988 issue of Computer magazine will be devoted to a wide range of topics in Neural Computing. Manuscripts that are either tutorial, survey, descriptive, case-study, applications-oriented or pedagogic in nature are immediately sought in the following areas: o Neural Network Architectures o Electronic and Optical Neurocomputers o Applications of Neural Networks in Vision, Speech Recognition and Synthesis, Robotics, Image Process- ing, and Learning o Self-Adaptive and Dynamically Reconfigurable Systems o Neural Network Models o Neural Algorithms and Models of Computation o Programming Neural Network Systems INSTRUCTIONS FOR SUBMITTING MANUSCRIPTS Manuscripts should be no more than 32-34 typewritten, double-spaced pages in length including all figures and ref- erences. No more than 12 references should be cited. Papers must not have been previously published nor currently sub- mitted for publication elsewhere. Manuscripts should have a title page that includes the title of the paper, full name of its author(s), affiliation(s), complete physical and electronic address(es), telephone number(s), a 200 word ab- stract, and a list of keywords that identify the central is- sues of the manuscript's content. DEADLINES o A 200 word abstract on the manuscript is due as soon as possible. o Eight (8) copies of the full manuscript is due by August 30, 1987. o Notification of acceptance is November 1, 1987. o Final version of the manuscript is due no later than December 1, 1987. SEND SUBMISSIONS AND QUESTIONS TO Bruce D. Shriver Editor-in-Chief, Computer IBM T. J. Watson Research Center P. O. Box 704 Yorktown Heights, NY 10598 Phone: (914) 789-7626 Electronic Mail Addresses: arpanet: shriver@ibm.com bitnet: shriver at yktvmh compmail+: b.shriver ------------------------------ Date: 5 Jun 87 18:14:46 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Smolensky on Connectionism: BBS Call for Commentators The following is the abstract of a forthcoming article on which BBS [Behavioral and Brain Sciences -- An international, interdisciplinary Journal of Open Peer Commentary, published by Cambridge University Press] invites self-nominations by potential commentators. (Please note that the editorial office must exercise selectivity among the nominations received so as to ensure a strong and balanced cross-specialty spectrum of eligible commentators. The procedure is explained after the abstract.) ----- On the Proper Treatment of Connectionism Paul Smolensky Institute of Cognitive Science University of Colorado Boulder CO 80309-0430 A set of hypotheses is formulated for a connectionist approach to cognitive modeling. These hypotheses are shown to be incompatible with the hypotheses embodied in traditional cognitive models. The connectionist models considered are massively parallel numerical com- putational systems that are a kind of continuous dynam- ical system. The numerical values in the system correspond semantically to fine-grained features below the level of the concepts used to describe the task domain. The level of analysis is intermediate between that of symbolic cognitive models and neural models. The explanations of behavior provided are like those in traditional physical sciences, unlike the explanations provided by symbolic models. Higher-level analyses of these connectionist models reveal subtle relations to symbolic models. Fundamen- tally parallel connectionist memory and linguistic processes are hypothesized to give rise to processes that are describable at a higher level as sequential rule application. At the lower level, computation has the character of massively parallel satisfaction of numerical constraints; at the higher level this can lead to competence characterizable by hard rules. Per- formance will typically deviate from competence since behavior is achieved not by interpreting hard rules but by satisfying soft constraints. The result is a picture in which traditional and connectionist theoretical con- structs collaborate intimately to provide an under- standing of cognition. ----- This is an experiment in using the Net to find eligible commentators for articles in the Behavioral and Brain Sciences (BBS), an international, interdisciplinary journal of "open peer commentary," published by Cambridge University Press, with its editorial office in Princeton NJ. The journal publishes important and controversial interdisciplinary articles in psychology, neuroscience, behavioral biology, cognitive science, artificial intelligence, linguistics and philosophy. Articles are rigorously refereed and, if accepted, are circulated to a large number of potential commentators around the world in the various specialties on which the article impinges. Their 1000-word commentaries are then co-published with the target article as well as the author's response to each. The commentaries consist of analyses, elaborations, complementary and supplementary data and theory, criticisms and cross-specialty syntheses. [...] Eligible individuals who judge that they would have a relevant commentary to contribute should contact the editor at the e-mail address indicated at the bottom of this message, or should write by normal mail to: Stevan Harnad Editor Behavioral and Brain Sciences 20 Nassau Street, Room 240 Princeton NJ 08542 (phone: 609-921-7771) Potential commentators should send their names, addresses, a description of their general qualifications and their basis for seeking to comment on this target article in particular to the address indicated earlier or to the following e-mail address: {seismo, psuvax1, bellcore, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.princeton.edu [Subscription information is available from Harry Florentine at Cambridge University Press: 800-221-4512] [Contact Harnad for further discussion of eligibility, application procedures, journal circulation, etc. -- KIL] -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 10 Jun 87 04:42:54 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Laming on Sensory Analysis: BBS Multiple Book Review The following is the abstract of a book that will be multiply reviewed in BBS [Behavioral and Brain Sciences -- An international, interdisciplinary Journal of Open Peer Commentary, published by Cambridge University Press]. Self-nominations by potential reviewers/commentators are invited. Please note that the editorial office must exercise selectivity among the nominations received so as to ensure a strong and balanced cross-specialty spectrum of eligible commentators. The procedure is explained after the abstract. ----- SENSORY ANALYSIS Donald Laming Department of Experimental Psychology University of Cambrdige Cambridge CB2 3EB ENGLAND ABSTRACT Sensory analysis is that initial, preconscious stage of perception at which features (edges, temporal discon- tinuities, and periodicities) are picked out from the random fluctuations that characterize the physical stimulation of sensory receptors. Sensory analysis may be studied by means of signal-detection, psychometric- function and threshold experiments, and my book, SEN- SORY ANALYSIS, presents a succinct, quasi-quantitative account of the phenomena revealed thereby. This account covers all five sensory modalities, emphasizing the similarities between them. A succinct account depends on identifying simple prin- ciples of wide generality, of which the most fundamen- tal are that (a) sensory discriminations are differen- tially coupled to the physical stimuli and that (b) small stimuli are subject to a square-law transform which makes them less detectable than they would other- wise be. These two principles are established by com- parisons between different configurations of two stimulus levels to be discriminated; they are realized within a simple physical-analogue model which affords certain low-level comparisons with neurophysiological observation. That physical-analogue model consists of a sequence of elementary operations on the stimulus con- stituting a stage of sensory processing. The concate- nation of two of three stages in cascade accommodates an increased range of experimental phenomena, espe- cially the detection of sinusoidal gratings. My BBS precis is organized in three parts: Part I sur- veys SENSORY ANALYSIS as economically as may be, begin- ning from the simplest, most fundamental ideas and working towards phenomena of increasing complexity. A rather short Part II reviews the most important alter- native models addressed to some part or other of the phenomena surveyed. Finally, a very short Part III con- tributes some metatheoretic remarks on the function of a theory of sensory discrimination. Potential commentators/reviewers should send their names, addresses, a description of their general qualifications and their basis for seeking to review this book in particular to the following USmail or Email address: Stevan Harnad, Editor Behavioral and Brain Sciences 20 Nassau Street, Room 240 Princeton NJ 08542 (phone: 609-921-7771) {seismo, psuvax1, bellcore, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.princeton.edu [See previous solicitations in AIList for the full blurb. -- KIL] -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: Fri, 12 Jun 87 15:40:35 edt From: John Grefenstette Subject: Conference - Genetic Algorithms Second International Conference on Genetic Algorithms and Their Applications July 28-31, 1987 MIT Cambridge, Massachusetts Sponsored By American Association for Artificial Intelligence Naval Research Laboratory Bolt Beranek and Newman, Inc. Genetic algorithms are adaptive search techniques based on principles derived from natural population genetics, and are currently being applied to a variety of difficult problems in science, engineering, and artificial intelligence. Topics for discussion will include: Fundamental research on genetic algorithms Machine learning using genetic algorithms Implementation techniques, especially on parallel processors Relationships to connectionism and other search and learning techniques Application of genetic algorithms Conference Committee: John H. Holland University of Michigan (Conference Chair) Lashon B. Booker Navy Center for Applied Research in AI Dave Davis Bolt Beranek and Newman, Inc. Kenneth A. De Jong George Mason University David E. Goldberg University of Alabama John J. Grefenstette Navy Center for Applied Research in AI (Program Chair) Stephen F. Smith Carnegie-Mellon Robotics Institute Stewart W. Wilson Rowland Institute for Science (Local Arrangements) The registration fee is $120 ($175 after June 15) and includes admission to all sessions, the Conference Proceedings, a Welcoming Reception, and all coffee breaks and lunches. The Conference Banquet is $30 additional per person. The Registration fee for students is $60. For registration forms and information concerning local arrangements, contact: Conference Services Office Room 7-111 Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, MA 02139 (617) 253-1703 For copies of the Conference Proceedings, contact: Lawrence Erlbaum Associates, Publishers 365 Broadway Hillsdale, New Jersey 07642 CONFERENCE PROGRAM TUESDAY, JULY 28, 1987 5:00 - 9:00 REGISTRATION 7:00 - 9:00 WELCOMING RECEPTION 7:00 - 9:00 TUTORIAL (if sufficient interest) WEDNESDAY, JULY 29, 1987 8:00 REGISTRATION 9:00 OPENING REMARKS 9:20 - 10:40 GENETIC SEARCH THEORY Finite Markov chain analysis of genetic algorithms David E. Goldberg and Philip Segrest An analysis of reproduction and crossover in a binary-coded genetic algorithm Clayton L. Bridges and David E. Goldberg Reducing bias and inefficiency in the selection algorithm James E. Baker Altruism in the bucket brigade Thomas H. Westerdale 10:40 - 11:00 COFFEE BREAK 11:00 - 12:00 ADAPTIVE SEARCH OPERATORS I Schema recombination in pattern recognition problems Irene Stadnyk An adaptive crossover distribution mechanism for genetic algorithms J. David Schaffer and Amy Morishima Genetic algorithms with sharing for multimodal function optimization David E. Goldberg and Jon Richardson 12:00 - 2:00 LUNCH 2:00 - 3:20 REPRESENTATION ISSUES The ARGOT strategy: adaptive representation genetic optimizer technique Craig G. Shaefer Nonstationary function optimization using genetic algorithms with dominance and diploidy David E. Goldberg and Robert E. Smith Genetic operators for high-level knowledge representations H. J. Antonisse and K. S. Keller Tree structured rules in genetic algorithms Arthur S. Bickel and Riva Wenig Bickel 3:20 - 3:40 COFFEE BREAK 3:40 - 5:00 KEYNOTE ADDRESS Genetic algorithms and classifier systems: foundations and future directions John H. Holland 7:00 - 9:00 BUSINESS MEETING THURSDAY, JULY 30, 1987 9:00 - 10:20 ADAPTIVE SEARCH OPERATORS II Greedy genetics G.E. Liepins, M.R. Hilliard, Mark Palmer and Michael Morrow Incorporating heuristic information into genetic search Jung Y. Suh and Dirk Van Gucht Using reproductive evaluation to improve genetic search and heuristic discovery Darrell Whitley Toward a unified thermodynamic genetic operator David J. Sirag and Paul T. Weisser 10:20 - 10:40 COFFEE BREAK 10:40 - 12:00 CONNECTIONISM AND PARALLELISM I Toward the evolution of symbols Charles P. Dolan and Michael G. Dyer SUPERGRAN: a connectionist approach to learning, integrating genetic algorithms and graph induction Deon G. Oosthuizen Parallel implementation of genetic algorithms in a classifier system George G. Robertson Punctuated equilibria: a parallel genetic algorithm J.P. Cohoon, S.U. Hegde, W.N. Martin and D. Richards 12:00 - 2:00 LUNCH 2:00 - 3:20 PARALLELISM II A parallel genetic algorithm Chrisila B. Pettey, Michael R. Leuze and John J. Grefenstette Genetic learning procedures in distributed environments Adrian V. Sannier II and Erik D. Goodman Parallelisation of probabilistic sequential search algorithms Prasanna Jog and Dirk Van Gucht Parallel genetic algorithms for a hypercube Reiko Tanese 3:20 - 3:40 COFFEE BREAK 3:40 - 5:00 CREDIT ASSIGNMENT AND LEARNING Bucket brigade performance: I. Long sequences of classifiers Rick L. Riolo Bucket brigade performance: II. Default hierarchies Rick L. Riolo Multilevel credit assignment in a genetic learning system John J. Grefenstette On using genetic algorithms to search program spaces Kenneth A. De Jong 6:30 - 10:00 CLAM BAKE FRIDAY, JULY 31, 1987 9:00 - 10:20 APPLICATIONS I A genetic system for learning models of consumer choice David Perry Greene and Stephen F. Smith A study of permutation crossover operators on the traveling salesman problem I.M. Oliver, D.J. Smith and J. R. C. Holland A classifier based system for discovering scheduling heuristics M.R. Hilliard, G.E. Liepins, Mark Palmer, Michael Morrow and Jon Richardson Using the genetic algorithm to generate LISP source code to solve the prisoner's dilemma Cory Fujiko and John Dickinson 10:20 - 10:40 COFFEE BREAK 10:40 - 12:00 APPLICATIONS II Optimal determination of user-oriented clusters: an application for the reproductive plan Vijay V. Raghavan and Brijesh Agarwal The genetic algorithm and biological development Stewart W. Wilson Genetic algorithms and communication link speed design: theoretical considerations Lawrence Davis and Susan Coombs Genetic algorithms and communication link speed design: constraints and operators Susan Coombs and Lawrence Davis 12:00 - 2:00 LUNCH 2:00 - 3:20 PANEL DISCUSSION: GA's and AI 3:20 - 3:40 COFFEE BREAK 3:40 - 5:00 INFORMAL DISCUSSION AND FAREWELL ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Mon Jun 15 03:16:29 1987 Date: Mon, 15 Jun 87 03:16:21 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #144 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Mon, 15 Jun 87 03:14 EDT Received: from relay.cs.net by RELAY.CS.NET id aa07555; 15 Jun 87 2:50 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10534; 15 Jun 87 2:50 EDT Date: Sun 14 Jun 1987 23:29-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #144 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 15 Jun 1987 Volume 5 : Issue 144 Today's Topics: Queries - Smalltalk-80 Implementations & AI Grad Schools & Machine Emotion Research & Neural Network Processors in High Technology & ML programming & ICOT Prolog, Theory - Complexity Theory, AI Tools - The ISI Grapher ---------------------------------------------------------------------- Date: 10 Jun 87 18:34:19 GMT From: ihnp4!alberta!sask!kusalik@ucbvax.Berkeley.EDU (Tony Kusalik) Subject: request for info on Smalltalk-80 implementations We are looking for a version of Smalltalk-80 for SUN-3's. We have contacted Berkeley about BSII, but the blurb that came back states "It [BSII] has not been updated to run on SUN 3 or to run under the X window system, although others have made these changes" Anyone know who these "others" might be? I.e. can anyone out there point me in the direction of a Smalltalk-80 system for SUN-3's? The Berkeley blurb mentions a SUN implementation done by L. Peter Deutsch and Allan M. Schiffman. Does anyone know of addresses (Email or snail-mail) for them? Tony Kusalik kusalik@sask.bitnet ...!{ihnp4,alberta}!sask!kusalik ------------------------------ Date: 8 Jun 87 13:59:44 GMT From: spe@SPICE.CS.CMU.EDU (Sean Engelson) Subject: AI grad schools? Can anyone give me any `inside' information on graduate CS-AI PhD programs? I know of a number of schools with such programs; I am interested in the opinions of people who have been involved in such programs, either as students or as professors. My main interests are in machine learning, analogical and common-sense reasoning, and natural language processing. Thank you, -- Credo, ergo absurdum est. LISP ::= ((())((Lots(())))(()(()(of(((Idiotic)())()()(Silly(()))()(Parentheses)))))) ---------------------------------------------------------------------- Sean Philip Engelson I have no opinions. Carnegie-Mellon University Therefore my employer is mine. Computer Science Department ---------------------------------------------------------------------- ARPA: spe@spice.cs.cmu.edu UUCP: {harvard | seismo | ucbvax}!spice.cs.cmu.edu!spe ------------------------------ Date: 12 Jun 87 02:03:39 GMT From: dartvax!uvm-gen!emerson.UUCP@seismo.css.gov (Tom "Oliver W. Jones" Emerson) Subject: Machine emotion research I would like information regarding emotion research in intelligent computers, including references if possible. If there is suffecient interest, I will report the contents of replies to the net. Thanx in advance, Tom E. ------------------------------ Date: 12 Jun 87 15:40:00 EDT From: LANTZ@RED.RUTGERS.EDU Subject: Neural network processors in High Technology The May issue of High Technology has an(other) article on neural networks. Would someone please send me the names and addresses of the four companies mentioned in the article. Thanks. Brian ------------------------------ Date: 9 Jun 87 19:36:37 GMT From: uunet!steinmetz!philabs!sbcs!sbstaff2!allen@seismo.css.gov ( Allen Leung) Subject: ML programming, anybody? Is there any one out there doing serious programming and/or research in ML (Meta Language)? I would like to hear from you. Don't trust me, I'm just an undergrad. Allen Leung, SUNY at Stony Brook. ------------------------------ Date: 11 Jun 87 18:12:40 GMT From: mit-vax!jouvelot@eddie.mit.edu (Pierre Jouvelot) Subject: Re: ML programming, anybody? In article <665@sbstaff2.UUCP> allen@sbstaff2.UUCP ( Allen Leung) writes: > > Is there any one out there doing serious programming and/or research > in ML (Meta Language)? I would like to hear from you. > Yes (I guess) !! I used it as an executable specification language for the semantic parallelization of imperative programs during my PhD research in France (I'm currently a PostDoc in MIT/LCS Programming Research Group). The overall program is about 3500 lines of ML code (with a few others in FranzLisp and MLYacc). The overall idea is described in my POPL'87 paper "Semantic Parallelization: A Practical Exercise in Abstract Interpretation" where both the theory (abstract interpretation) and practice (use of ML) are introduced (for courageous people, there is also my PhD thesis ... written in french :-) Note that I used the Cambridge/INRIA Version which is older and slightly different from SML. The main problem I had was related to the lack of "real" separate-compilation facility. This should disappear with newer versions that introduce modules. Besides this, ML is a very fine language which should have a more widespread use. Pierre -- Pierre Jouvelot Room NE43-403 ARPA: jouvelot@xx.lcs.mit.edu Lab for Computer Science USENET: decvax!mit-vax!jouvelot MIT (or mcvax!litp!pj) 545, Technology Square Cambridge, MA 02139 USA ------------------------------ Date: Thu, 11 Jun 87 08:31:20 EDT From: elsaesser%mwcamis@mitre.arpa Subject: Say, what ever happened to ... ICOT Prolog????? It seems ages ago that the 5th generation project was going to reinvent AI in a Prolog "engine" that was to do 10 gazillion " LIPS". Anyone know what happened? I mean, if you can make so many "quality" cars (sans auto transmission, useful A/C, paint that can take rain and sun, etc.), why can't you make a computer that runs an NP-complete applications language in real time??? Simi-seriously, what is the status of the 5th generation project, anyone got an update? chris (elsaesser%mwcamis@mitre.arpa) [See the June IEEE Spectrum, "Next-Generation Race Bogs Down", Karen Fitzgerald and Raul Wallich, pp. 28-33, for a review. The Japanese effort is doing well enough in its parallel architecture development and is making some progress in "knowledge programming", but has dropped VLSI technology and made little headway in AI and knowledge representation. Competitive efforts in the U.S. and Europe have also had the most success in hardware. The real question now is whether the 5th-generation push has given Japan the kind of computer-science infrastructure that it needs to compete and perhaps pull out ahead in algorithm development. My guess is that it has not (because the software part of the effort was too small). An interesting sign of change, though, is the Japanese government's invitation to Western universities to set up branches in Japan. I assume that Japanese leaders will always come from Tyodai or Kyodai, but perhaps computer scientists will be educated in a different tradition. -- KIL] ------------------------------ Date: 10 Jun 87 08:33:39 GMT From: mcvax!botter!klipper!biep@seismo.css.gov (J. A. "Biep" Durieux) Subject: Re: What philosophical problems does complexity theory yield? In article <789@klipper.cs.vu.nl> biep@cs.vu.nl (J. A. "Biep" Durieux) writes: >But, isn't anything which cannot be turned into a constant-time process >philosophically annoying? Why just hassling about non-polynomial time >solutions? Am I missing something? In article <2258@cvl.umd.edu> ramesh@cvl.UUCP (Ramesh Sitaraman) writes: >Yes, you are missing the point !! > >The difference between a polynomial and non-polynomial solution for >a problem is the difference between structure and a complete lack >of it. Thanks a lot, this sounds much more relevant than just computation time. But, isn't finding the smallest element of a set solvable only by "dumb exhaustive search" either? Are people having that much trouble with such a linear algorithm too? Also, thanks for including the defs! Such things make the net a whole lot more readable. But: please don't put your mail address on the "Follow-up-to: " line. I'm having a terrible time getting this article out! Inews feeding time ------------------------------ Date: Sat, 13 Jun 87 13:35:12 PDT From: Gabriel Robins Subject: The ISI Grapher Greetings, Due to the considerable interest drawn by the ISI Grapher so far, I am posting this abstract summarizing its function and current status. Interested parties may obtain further information by directly sending EMail to "gabriel@vaxa.isi.edu" or by writing to: Gabriel Robins Intelligent Systems Division Information Sciences Institute 4676 Admiralty Way Marina Del Rey, Ca 90292-6695 If you want documentation in hardcopy, please include your U.S. Mail address. Gabe ---------------------------------------------------------------------- The ISI Grapher June, 1987 Gabriel Robins Intelligent Systems Division Information Sciences Institute The ISI Grapher is a set of functions that convert an arbitrary graph structure (or relation) into an equivalent pictorial representation and displays the resulting diagram. Nodes and edges in the graph become boxes and lines on the workstation screen, and the user may then interact with the Grapher in various ways via the mouse and the keyboard. The fundamental motivation which gave birth to the ISI Grapher is the observation that graphs are very basic and common structures, and the belief that the ability to quickly display, manipulate, and browse through graphs may greatly enhance the productivity of a researcher, both quantitatively and qualitatively. This seems especially true in knowledge representation and natural language research. The ISI Grapher is both powerful and versatile, allowing an application-builder to easily build other tools on top of it. The ISI NIKL Browser is an example of one such tool. The salient features of the ISI Grapher are its portability, speed, versatility, and extensibility. Several additional applications were already built on top of the ISI Grapher, providing the ability to graph lists, flavors, packages, divisors, functions, and Common-Loops classes. Several basic Grapher operations may be user-controlled via the specification of alternate functions for performing these tasks. These operations include the drawing of nodes and edges, the selection of fonts, the determination of print-names, pretty-printing, and highlighting operations. Standard definitions are already provided for these operations and are used by default if the application-builder does not override them by specifying his own custom-tailored functions for performing the same tasks. The ISI Grapher now spans about 100 pages of CommonLisp code. The 120-page ISI Grapher manual is available; this manual describes the general ideas, the interface, the application-builder's back-end, the algorithms, the implementation, and the data structures. The ISI Grapher presently runs on both Symbolics (6 & 7) and TI Explorer workstations. If you are interested in more information, the sources themselves, or just the documentation/manual, please feel free to forward your U.S. Mail address to "gabriel@vaxa.isi.edu" or write to "Gabriel Robins, c/o Information Sciences Institute, 4676 Admiralty Way, Marina Del Rey, Ca 90292-6695." ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 16 03:21:33 1987 Date: Tue, 16 Jun 87 03:21:20 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #145 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 16 Jun 87 03:09 EDT Received: from relay.cs.net by RELAY.CS.NET id ab07788; 15 Jun 87 3:13 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10641; 15 Jun 87 3:11 EDT Date: Sun 14 Jun 1987 23:45-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #145 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 15 Jun 1987 Volume 5 : Issue 145 Today's Topics: Query - Why Did The $6,000,000 Man Run So Slowly?, Science - Applying AI Models to Biology ---------------------------------------------------------------------- Date: Fri, 12 Jun 87 00:51:41 EDT From: tim@linc.cis.upenn.edu (Tim Finin) Subject: why did the $6,000,000 man run so slowly? Why did the six million dollar man run so slowly? Some time ago, Pat Hays posted a message in which he asked people for explanations for the fact that Dr. Who's tardis is bigger on the inside than it appears to be from the outside. He was trying, of course, to discover something about our common sense model of the physical world. I have a similar question which might shed some light on our common sense notions of time and actions: why did the six million dollar man run so slowly? As you recall, the six million dollar man (from the popular TV show in the early '70's) had bionic legs which enabled him to run at super-human speeds. However, when the producers wanted to show him doing this, they slowed down the image of him running. That is, to depict him running at incredibly fast speeds, they showed an image of him moving in "slow motion". Id like to collect explanations for this fact. Tim. ------------------------------ Date: 12 Jun 87 20:51:51 GMT From: ihnp4!homxb!houxm!hou2d!avr@ucbvax.Berkeley.EDU (Adam V. Reed) Subject: Re: Why did the six-million dollar man run so slowly? Slow motion is commonly used in TV (and before that, newsreel) reports to represent very fast motion (e.g. in horse races and other sports events). My guess is that this originated through use of free "photo finish" footage, originally filmed for the use of sport-event judges, in early movie newsreels. If my guess is right, the representation of fast movement with slow-motion footage uses a learned but highly familiar mental association. Adam Reed (hou2d!adam) ------------------------------ Date: 13 Jun 87 03:16:03 GMT From: code@sphinx.uchicago.edu (paul robinson wilson) Subject: Re: Why did the six-million dollar man run so slowly? In article <1431@hou2d.UUCP> avr@hou2d.UUCP (Adam V. Reed) writes: >Slow motion is commonly used in TV (and before that, newsreel) reports >to represent very fast motion (e.g. in horse races and other sports >events). My guess is that this originated through use of free "photo >finish" footage, originally filmed for the use of sport-event judges, >in early movie newsreels. If my guess is right, the representation of >fast movement with slow-motion footage uses a learned but highly >familiar mental association. I think it may be more subtle than that. There is a general tendency for effective, competent motion to be smooth and for large motions to be relatively slow. A long-legged runner runs more "slowly" than a short-legged one, but covers more ground. A jaguar moves fluidly and less hurriedly than its usual prey, making large bounds seemingly effortlessly. By contrast, the little kid trying to keep up with the big kids moves its legs very fast. Naturally, if we saw speeded-up film of the $ 6 Meg man, we'd think he looked comical, with his legs moving very rapidly, like a small (impotent) creature's. Slow-motion, however, looks smooth and graceful, revealing the grace with which we all move, but seldom notice. Our ability to appreciate this (intended) effect without the accompanying (unintended) impression of his moving quite slowly, however, may in fact depend on our "being used to it" from television sports, etc. We appreciate the obvious grace while suspending our judgement about speed. The _right_ way to show it, I guess, would have been to have Lee Majors bound 20 ft. (or thereabouts) at a time, and quickly. Besides being a bit difficult to accomplish, it's also a little hard on the skeletal structure. They would have gone through stuntmen at quite a clip :-). (By the way, I believe Lee Majors is a rather short guy, and would have looked especially comical in sped-up film, coveing significant ground, with normal stuff to gauge him against.) | Paul R. Wilson ph.: (312) 947-0740 uucp: ...!ihnp4!uicbert!wilson | | Electronic Mind Control Lab if no answer: ...ihnp4!gargoyle!sphinx!code | | UIC EECS Dept. (M/C 154) arpa: uicbert!wilson@uxc.cso.uiuc.edu | | P.O.Box 4348 Chicago,IL 60680 | ------------------------------ Date: 13 Jun 87 06:18:52 GMT From: pattis@june.cs.washington.edu (Richard Pattis) Subject: Re: Why did the six-million dollar man run so slowly? I've thought that the slowdown was not from the perspective of the viewer, but from the perspective of the the $6M man. The viewer, viewing from the frame of the $6M man, is moving so fast that everything else seems slowed down. ------------------------------ Date: 13 Jun 87 17:49:31 GMT From: super.upenn.edu!linc.cis.upenn.edu!mayerk@RUTGERS.EDU (Kenneth Mayer) Subject: Re: Why did the six-million dollar man run so slowly? Occaisionally, the producers _did_ show Lee Majors in a speeded up shot. The effect was comical. (As I recall, there was this old farmer watching from the porch of his house as Mr. $6million sprinted across his field.) I like the cougar metaphor. Wildlife films of such an animal in normal speed are choppy, incredibly brief, and usuall ends with the felling of the prey. In slow-mo we get a chance to see the beautiful detail of the predator flying by. >From a cinematic viewpoint, the camera director/special effects director had to do something to show that Steve Austin wasn't simply jogging across a field like the rest of us. Slowing the file speed (and speeding up apparent time) looks comical, like an old Keystone Cops film. Stretching out the time line increases tension. The viewer gets a chance to examine more detail per sec. of real time. Exactly the way a novel will be incredibly brief during transitions, and excrutiatingly deatailed during climaxes. (I just finished reading Misery, by Stephen King. For a good reflective look at a writer's art, packaged in a really good thriller, borrow this book from the library for a summer weekend reader.) Kenneth Mayer mayerk@eniac.seas.upenn.edu ------------------------------ Date: 10 Jun 87 09:33:34 GMT From: nosc!humu!uhccux!todd@sdcsvax.ucsd.edu (The Perplexed Wiz) Subject: Re: Taking AI models and applying them to biology... In article <836@pixar.UUCP> davel@pixar.UUCP (David Longerbeam) writes: >In article <622@unicus.UUCP>, craig@unicus.UUCP (Craig D. Hubley) writes: >> This description of the human memory system, though cloaked in vaguer terms, >> corresponds more or less one-to-one with the traditional computer >> architecture we all know and love. To wit: > [description deleted] >> At least this far, this theory appears to owe a lot to computer science. >> Granted, there is lots of empirical evidence in favour, but we all know >> how a little evidence can go far too far towards developing an analogy. >One of my philosophy professors in college offered the observation that >models for the human mind have always seemed to correspond to the most >advanced form of technology at that given point in history. He could It's true that theories of cognition often reflect the current popular technology. But before we start arguing current theories as reflections of computer science and physiology, I suggest we at least have some common starting point for our discussion. I don't want to suggest that you need a Ph.D. in Cognitive Psychology to discuss the subject, but you might want to consider reading one of the many intro texts on the subject before leaping to any speculations (wild or otherwise :-). An intro text I often recommend to people with a more than casual interest in cognition is: Anderson, John (1985). Cognitive Psychology and Its Implications. (2nd edition) New York: W.H. Freeman and Co. [The 1st edition also has much to recommend it. It was written from a psychological viewpoint, and introduces vocabulary and concepts that may be unfamiliar to computer scientists. The 2nd edition was rewritten with an AI (or cognitive psychology!) vocabulary, hence risks echoing the preconceptions of the field instead of contributing fresh insights. -- KIL] If you are interested in a historical perspective of psychological research, I suggest you take a peek at: Hearst, Eliot (Ed.) (1979). The First Century of Experimental Psychology. Hillsdale, New Jersey: Lawrence Erlbaum Associates, Pub. And finally, though I don't always agree with what Richard Gregory has to say, I always enjoy hearing or reading his ideas and theories. His "Mind in Science" is an interesting speculative book. Gregory, Richard (1981). Mind in Science: A History of Explanations in Psychology and Physics. Cambridge: Cambridge University Press Well, I hope we at least have some common reference point now... Todd Ogasawara "With a good wind behind me and and a lot of luck... Ph.D. in Psychology later this year :-)" -- Todd Ogasawara, U. of Hawaii Computing Center UUCP: {ihnp4,seismo,ucbvax,dcdwest}!sdcsvax!nosc!uhccux!todd ARPA: uhccux!todd@nosc.MIL INTERNET: todd@uhccux.UHCC.HAWAII.EDU ------------------------------ Date: Wed, 10 Jun 87 09:51 EDT From: Seth Steinberg Subject: Borrowing from Biology [Half in Jest] Actually, the biologists have been borrowing from the history of the Roman Empire. Cincinatus comes down from his farm and codifies the laws for the Republic and creates a nearly perfect mechanism which starts taking over the Mediterranean basin. By providing for a means of succession (read "DNA replication"), the Empire is able to achieve higher levels of organization. Unfortunately, the military (read "the immune system"), slowly grows in strength as the Empire expands and finally reaches a limit to its expansion and spends the next millenium rotting away in Byzantium. Theories about entropy are about complex systems in general, not just the behavior of energy in steam engines. Biologists have latched onto them to account for aging in organisms and to explain the epochs of evolution. (Why aren't there any new phyla being created?) If you've ever tried to make a major change in a decade old program think of what the biologists are up against with their billion year old kludges. Last month, an article in Scientific American described a glucose complex based aging mechanism, arguing that many aging effects could be caused by very slow chemical reactions induced by the operating environment. Next month we may discover an actual internal counter within each cell. It is quite probable that there are dozens of mechanisms at work. With 90% of the genome encoding for garbage, elegant design is more of a serendipity than the norm. Seth Steinberg sas@bbn.com P.S. Did you notice the latest kludge? They've found a gene whose DNA complement also encodes a gene! Kind of like a 68000 program you can execute if you put a logical complement on each instruction fetch. Neat, huh? ------------------------------ Date: 12 Jun 87 16:08:04 GMT From: hao!boulder!eddy@ames.arpa (Sean Eddy) Subject: Re: Taking AI models and applying them to biology... In article <1331@sigi.Colorado.EDU> pell@boulder.Colorado.EDU writes: >It would seem to me that the step that is likely to give the cell trouble >is not mitosis but DNA replication. If a whole chromosome lost or >non-disjoined, that cell is in some serious trouble. Progressive >accumulation mistakes through replication and general maintanence seems a more >likely culprit. "General maintenance" is a very important thing to bring up. It seems to me that replication/mitosis can't be the whole story in aging. One must also propose other models because there are cells that do not divide after a certain point, yet still age and die. Neurons are the classic example; not only do they not divide, they cannot even be replaced (in humans) if damaged. - Sean Eddy - MCD Biology; U. of Colorado at Boulder; Boulder CO 80309 - eddy@boulder.colorado.EDU !{hao,nbires}!boulder!eddy - - "So what. Big deal." - - Emilio Lazardo ------------------------------ Date: 13 Jun 87 23:03:16 GMT From: mcvax!lambert@seismo.css.gov (Lambert Meertens) Subject: Re: Taking AI models and applying them to biology... In article <836@pixar.UUCP> davel@pixar.UUCP (David Longerbeam) writes: > In article <622@unicus.UUCP>, craig@unicus.UUCP (Craig D. Hubley) writes: | | > This description of the human memory system, though cloaked in vaguer terms, | > corresponds more or less one-to-one with the traditional computer | > architecture we all know and love. To wit: | | [description deleted] | | > At least this far, this theory appears to owe a lot to computer science. | > Granted, there is lots of empirical evidence in favour, but we all know | > how a little evidence can go far too far towards developing an analogy. | | One of my philosophy professors in college offered the observation that | models for the human mind have always seemed to correspond to the most > advanced form of technology at that given point in history. I find the connection between models of human memory as developed in cognitive psychology and existing computer architectures rather tenuous. The main similarity appears to be that several levels of memory can be discerned, but the suggested analogy in function is a bit far-fetched. It is perhaps worth pointing out that much of the current models in cognitive psychology can already be found in the pioneering work of Otto Selz (Muenchen, 1881 - Auschwitz, 1943), antedating the computer era. -- Lambert Meertens, CWI, Amsterdam; lambert@cwi.nl ------------------------------ Date: Thu, 11 Jun 87 13:48:05 BST From: Graham Higgins Subject: Re: Taking AI models and applying them to biology... In article <622@unicus.UUCP>, craig@unicus.UUCP (Craig D. Hubley) writes: > I was semi-surprised in recent months to discover that cognitive psychology, > far from developing a bold new metaphor for human thinking, has (to a degree) > copied at least one metaphor from third-generation computer science. Psychology freely borrows *any* models that will help it get a grip on characterising and explaining the phenomena of cognition. Over the years, analogies of the workings of the mind have been constructed from : windmills, hydraulic systems, telephone switching exchanges and latterly, the computer (or more properly, information-processing devices). The one thing that all these analogies have in common is that they draw on the technological state-of-the-art of the time. (The "internal combustion engine" analogy is a new one to me). David Longerbeam's comment about the requirement for empiricism is valid in this instance. Donald Hebb assumed a separation of STM and LTM in a 1949 paper (and that's going back quite some time, only a year after Shockley's invention of the transistor). It is unlikely that the computer-architecture construct of "archived storage" played any part in Hebb's dichotomising of human memory. It appears that this is one example of a model developed within cognitive psychology, independently of developments in computer architecture. (I'm not well-versed in comp.sci. history - but it seems reasonable to conjecture that Hebb was unaware of the notions of "archived storage" when he was developing his dichotomisation). > This description of the human memory system, though cloaked in vaguer terms, > corresponds more or less one-to-one with the traditional computer > architecture we all know and love ... > > - senses have "iconic" and "echo" memories analogous to buffers. > - short term memory holds information that is organized for quick > processing, much like main storage in a computing system. > - long term memory holds information in a sort of semantic > association network where large related pieces of information > reside, similar to backing or "archived" computing storage. I think that this is somewhat of an over-simplification. There are quite a few phenomena arising from studies of "iconic", "echoic", "short-term" and "long-term" areas of human memory which do not fit so tamely into a computer-architecture model. Thus, there has *not* been uncritical acceptance of either that the "iconic" and "echoic" aspects of memory are passive or that memory can be simply dichotomised into into STM and LTM sections. In the absence of anything better, the analogies will do for now, but there are too many phenomena which don't fit in to these analogies for them to anything but convenient for the moment. One of the disciplinary traits actively promoted in psychology (be it cognitive, social, experimental, etc.) is a high degree of circumspection. (There is a tradition that one never sees a one-armed psychologist - "on one hand .... and on the other ... "). Thus models and analogies *can* be freely borrowed from other areas and exploited for what they offer, for as long as they exhibit some level of descriptive utility. It is instructive to note that contemporary cognitive psychologists no longer use windmills or telephone exchanges (or even the internal combustion engine) as analogies of the workings of the mind. These particular analogies have outlived their usefulness and have been discarded (I hope!). Graham Higgins || The opinions expressed above Hewlett-Packard Labs || are not to be contrued as the Bristol, U.K. || opinions, stated or otherwise, gjh@hplb.csnet +44 272 799910 xt 4060 || of Hewlett-Packard ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Wed Jun 17 03:52:17 1987 Date: Wed, 17 Jun 87 03:52:07 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #146 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Wed, 17 Jun 87 03:13 EDT Received: from relay.cs.net by RELAY.CS.NET id ae14738; 16 Jun 87 2:55 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10779; 16 Jun 87 2:55 EDT Date: Mon 15 Jun 1987 23:20-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #146 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 16 Jun 1987 Volume 5 : Issue 146 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 9 Jun 87 22:12:32 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In article <812@mind.UUCP> Stevan Harnad replies: With regard physical invertibility and the A/D distinction: > >> a *digitized* image -- in your terms... is "analog" in the >> information it preserves and not in the information lost. This >> seems to me to be a very unhappy choice of terminology! > > For the time being, I've acknowledged that >my invertibility criterion is, if not necessarily unhappy, somewhat >surprising in its implications, for it implies (1) that being analog >may be a matter of degree (i.e., degree of invertibility) and (2) even >a classical digital system must be regarded as analog to a degree ... Grumble. These consequences only *seem* surprising if we forget that you've redefined "analog" in a non-standard manner; this is precisely I why I keep harping on your terminology. Compare them with what you're really saying: "physical invertibility is a matter of degree" or "a classical digital system still employs physically invertible representations" -- both quite humdrum. With regard to the symbolic AI approach to the "symbol-grounding problem": > >One proposal, as you note, is that a pure symbol-manipulating system can be >"grounded" by merely hooking it up causally in the "right way" to the outside >world with simple (modular) transducers and effectors. ... I have argued >that [this approach] simply won't succeed in the long run (i.e., as we >attempt to approach an asymptote of total human performance capacity ...) >...In (1) a "toy" case ... the right causal connections could be wired >according to the human encryption/decryption scheme: Inputs and outputs could >be wired into their appropriate symbolic descriptions. ... But none but the >most diehard symbolic functionalist would want to argue that such a simple >toy model was "thinking," ... The reason is that we are capable of >doing *so much more* -- and not by an assemblage of endless independent >modules of essentially the same sort as these toy models, but by some sort of >(2) integrated internal system. Could that "total" system be just an >oversized toy model -- a symbol system with its interpretations "fixed" by a >means analogous to these toy cases? I am conjecturing that it is not. I think your reply may misunderstand the point of my objection. I'm not trying to defend the intentionality of "toy" programs. I'm not even particularly concerned to *defend* the symbolic approach to AI (I personally don't even believe in it). I'm merely trying to determine exactly what your argument against symbolic AI is. I had thought, perhaps wrongly, that you were claiming that the interpretations of systems conceived by symbolic AI system must somehow inevitably fail to be "grounded", and that only a system which employed "analog" processing in the way you suggest would have the causal basis required for fixing an interpretation. In response, I pointed out first that advocates of the symbolic approach already understand that causal commerce with the environment is necessary for intentionality: they envision the use of complex perceptual systems to provide the requisite "grounding". So it's not as though the symbolic approach is indifferent to this issue. And your remarks against "toy" systems and "hard-wiring" the interpretations of the inputs are plain unfair -- the symbolic approach doesn't belittle the importance or complexity of what perceptual systems must be able to do. It is in total agreement with you that a truly intentional system must be capable of complex adaptive performance via the use of its sensory input -- it just hypothesizes that symbolic processing is sufficient to achieve this. And, as I tried to point out, there is just no reason that a modular, all-digital system of the kind envisioned by the symbolic approach could not be entirely "grounded" BY YOUR OWN THEORY OF "GROUNDEDNESS": it could employ "physically inevertible" representations (only they would be digital ones), from these it could induct reliable "feature filters" based on training (only these would use digital rather than analog techniques), etc. I concluded that the symbolic approach appears to handle your so-called "grounding problem" every bit as well as any other method. Now comes the reply that you are merely conjecturing that analog processing may be required to realize the full range of human, as opposed to "toy", performance -- in short, you think the symbolic approach just won't work. But this is a completely different issue! It has nothing to do with some mythical "symbol grounding" problem, at least as I understand it. It's just the same old "intelligent-behavior-generating" problem which everyone in AI, regardless of paradigm, is looking to solve. >From this reply, it seems to me that this alleged "symbol-grounding problem" is a real red-herring (it misled me, at least). All you're saying is that you suspect that mainstream AI's symbol system hypothesis is false, based on its lack of conspicuous performance-generating sucesses. Obviously everyone must recognize that this is a possibility -- the premise of symbolic AI is, after all, only a hypothesis. But I find this a much less interesting claim than I originally thought -- conjectures, after all, are cheap. It *would* be interesting if you could show, as, say, the connectionist program is trying to, how analog processing can work wonders that symbol-manipulation can't. But this would require detailed research, not speculation. Until then, it remains a mystery why your proposed approach should be regarded as any more promising than any other. Anders Weinstein BBN Labs ------------------------------ Date: 10 Jun 87 21:28:23 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., Cambridge, MA writes: > There's no [symbol] grounding problem, just the old > behavior-generating problem Before responding to the supporting arguments for this conclusion, let me restate the matter in what I consider to be the right way. There is: (1) the behavior-generating problem (what I have referred to as the problem of devising a candidate that will pass the Total Turing Test), (2) the symbol-grounding problem (the problem of how to make formal symbols intrinsically meaningful, independent of our interpretations), and (3) the conjecture (based on the existing empirical evidence and on logical and methodological considerations) that (2) is responsible for the failure of the top-down symbolic approach to solve (1). >>my [SH's] invertibility criterion is, if not necessarily unhappy, somewhat >>surprising in its implications, for it implies that (1) being analog may >>be a matter of degree (i.e., degree of invertibility) and that (2) even >>a classical digital system must be regarded as analog to a degree ... > > These consequences only *seem* surprising if we forget that you've > redefined "analog" in a non-standard manner... you're really saying: > "physical invertibility is a matter of degree" or "a classical digital > system still employs physically invertible representations" -- both > quite humdrum. You've bypassed the three points I brought up in replying to your challenge to my invertibility criterion for an analog transform the last time: (1) the quantization in standard A/D is noninvertible, (2) a representation can only be analog in what it preserves, not in what it fails to preserve, and, in cognition at any rate, (3) the physical shape of the signal may be what matters, not the "message" it "carries." Add to this the surprising logical consequence that a "dedicated" digital system (hardwired to its peripherals) would be "analog" in its invertible inputs and outputs according to my invertibility criterion, and you have a coherent distinction that conforms well to some features of the classical A/D distinction, but that may prove to diverge, as I acknowledged, sufficiently to make it an independent, "non-standard" distinction, unique to cognition and neurobiology. Would it be surprising if classical electrical engineering concepts did not turn out to be just right for mind-modeling? > I [AW] had thought, perhaps wrongly, that you were claiming that the > interpretations of systems conceived by symbolic AI system must somehow > inevitably fail to be "grounded", and that only a system which employed > "analog" processing in the way you suggest would have the causal basis > required for fixing an interpretation. That is indeed what I'm claiming (although you've completely omitted the role of the categorical representations, which are just as critical to my scheme, as described in the CP book). But do make sure you keep my "non-standard" definition of analog in mind, and recall that I'm talking about asymptotic, human-scale performance, not toy systems. Toy systems are trivially "groundable" (even by my definition of "analog") by hard-wiring them into a dedicated input/output system. But the problem of intrinsic meaningfulness does not arise for toy models, only for devices that can pass the Total Turing Test (TTT). [The argument here involves showing that to attribute intentionality to devices that exhibit sub-TTT performance is not justified in the first place.] The conjecture is accordingly that the modular solution (i.e., hardwiring an autonomous top-down symbolic module to conventional peripheral modules -- transducers and effectors) will simply not succeed in producing a candidate that will be able to pass the Total Turing Test, and that the fault lies with the autonomy (or modularity) of the symbolic module. But I am not simply proposing an unexplicated "analog" solution to the grounding problem either, for note that a dedicated modular system *would* be analog according to my invertibility criterion! The conjecture is that such a modular solution would not be able to meet the TTT performance criterion, and the grounds for the conjecture are partly inductive (extrapolating symbolic AI's performance failures), partly logical and methodological (the grounding problem), and partly theory and data-driven (psychophysical findings in human categorical perception). My proposal is not that some undifferentiated, non-standard "analog" processing must be going on. I am advocating a specific hybrid bottom-up, symbolic/nonsymbolic rival to the pure top-down symbolic approach (whether or not the latter is wedded to peripheral modules), as described in the volume under discussion ("Categorical Perception: The Groundwork of Cognition," CUP 1987). > advocates of the symbolic approach already understand that causal > commerce with the environment is necessary for intentionality: they > envision the use of complex perceptual systems to provide the > requisite "grounding". So it's not as though the symbolic approach > is indifferent to this issue. This is the pious hope of the "top-down" approach: That suitably "complex" perceptual systems will meet for a successful "hook-up" somewhere in the middle. But simply reiterating it does not mean it will be realized. The evidence to date suggests the opposite: That the top-down approach will just generate more special-purpose toys, not a general purpose, TTT-scale model of human performance capacity. Nor is there any theory at all of what the requisite perceptual "complexity" might be: The stereotype is still standard transducers that go from physical energy via A/D conversion straight into symbols. Nor does "causal commerce" say anything: It leaves open anything from the modular symbol-cruncher/transducer hookups of the kind that so far only seem capable of generating toy models, to hybrid, nonmodular, bottom-up models of the sort I would advocate. Perhaps it's in the specific nature of the bottom-up grounding that the nature of the requisite "complexity" and "causality" will be cashed in. > your remarks against "toy" systems and "hard-wiring" the > interpretations of the inputs are plain unfair -- the symbolic > approach doesn't belittle the importance or complexity of what > perceptual systems must be able to do. It is in total agreement > with you that a truly intentional system must be capable of complex > adaptive performance via the use of its sensory input -- it just > hypothesizes that symbolic processing is sufficient to achieve this. And I just hypothesize that it isn't. And I try to say why not (the grounding problem and modularity) and what to do about it (bottom-up, nonmodular grounding of symbolic representations in iconic and categorical representations). > there is just no reason that a modular, all-digital system of the > kind envisioned by the symbolic approach could not be entirely > "grounded" BY YOUR OWN THEORY OF "GROUNDEDNESS": it could employ > "physically inevertible" representations (only they would be digital > ones), from these it could induct reliable "feature filters" based on > training (only these would use digital rather than analog techniques), > etc. ... the symbolic approach appears to handle your so-called > "grounding problem" every bit as well as any other method. First of all, as I indicated earlier, a dedicated top-down symbol-crunching module hooked to peripherals would indeed be "grounded" in my sense -- if it had TTT-performance power. Nor is it *logically impossible* that such a system could exist. But it certainly does not look likely on the evidence. I think some of the reasons we were led (wrongly) to expect it were the following: (1) The original successes of symbolic AI in generating intelligent performance: The initial rule-based, knowledge-driven toys were great successes, compared to the alternatives (which, apart from some limited feats of Perceptrons, were nonexistent). But now, after a generation of toys that show no signs of converging on general principles and growing up to TTT-size, the inductive evidence is pointing in the other direction: More ad hoc toys is all we have grounds to expect. (2) Symbol strings seemed such hopeful candidates for capturing mental phenomena such as thoughts, knowledge, beliefs. Symbolic function seemed like such a natural, distinct, nonphysical level for capturing the mind. Easy come, easy go. (3) We were persuaded by the power of computation -- Turing equivalence and all that -- to suppose that computation (symbol-crunching) just might *be* cognition. If every (discrete) thing anyone or anything (including the mind) does is computationally simulable, then maybe the computational functions capture the mental functions? But the fact that something is computationally simulable does not entail that it is implemented computationally (any more than behavior that is *describable* as ruleful is necessarily following an explicit rule). And some functions (such as transduction and causality) cannot be implemented computationally at all. (4) We were similarly persuaded by the power of digital coding -- the fact that it can approximate analog coding as closely as we please (and physics permits) -- to suppose that digital representations were the only ones we needed to think about. But the fact that a digital approximation is always possible does not entail that it is always practical or optimal, nor that it is the one that is actually being *used* (by, say, the brain). Some form of functionalism is probably right, but it certainly need not be symbolic functionalism, or a functionalism that is indifferent to whether a mental function or representation is analog or digital: The type of implementation may matter, both to the practical empirical problem of successfully generating performance and to the untestable phenomenological problem of capturing qualitative subjective experience. And some functions (let me again add), such as transduction and (continuous) A/A, cannot be implemented purely symbolically at all. A good example to bear in mind is Shepard's mental rotation experiments. On the face of it, the data seemed to suggest that subjects were doing analog processing: In making same/different judgments of pairs of successively presented 2-dimensional projections of 3-dimensional, computer-generated, unfamiliar forms, subjects' reaction times for saying "same" when one stimulus was in a standard orientation and the other was rotated were proportional to the degree of rotation. The diehard symbolists pointed out (correctly) that the proportionality, instead of being due to the real-time analog rotation of a mental icon, could have been produced by, say, (1) serially searching through the coordinates of a digital grid on which the stimuli were represented, with more distant numbers taking more incremental steps to reach, or by (2) doing inferences on formal descriptions that became more complex (and hence time-consuming) as the orientation became more eccentric. The point, though, is that although digital/symbolic representations were indeed possible, so were analog ones, and here the latter would certainly seem to be more practical and parsimonious. And the fact of the matter -- namely, which kinds of representations were *actually* used -- is certainly not settled by pointing out that digital representations are always *possible.* Maybe a completely digital mind would have required a head the size of New York State and polynomial evolutionary time in order to come into existence -- who knows? Not to mention that it still couldn't do the "A" in the A/D... > [you] reply that you are merely conjecturing that analog processing > may be required to realize the full range of human, as opposed to "toy", > performance -- in short, you think the symbolic approach just won't > work. But this... has nothing to do with some mythical "symbol > grounding" problem, at least as I understand it. It's just > the same old "intelligent-behavior-generating" problem which everyone > in AI, regardless of paradigm, is looking to solve... All you're > saying is that you suspect that mainstream AI's symbol system > hypothesis is false, based on its lack of conspicuous > performance-generating successes. Obviously everyone must recognize > that this is a possibility -- the premise of symbolic AI is, after > all, only a hypothesis. I'm not just saying I think the symbolic hypothesis is false. I'm saying why I think it's false (ungroundedness) and I'm suggesting an alternative (a bottom-up hybrid). > But I find this a much less interesting claim than I originally > thought -- conjectures, after all, are cheap. It *would* be > interesting if you could show, as, say, the connectionist program > is trying to, how analog processing can work wonders that > symbol-manipulation can't. But this would require detailed research, > not speculation. Until then, it remains a mystery why your proposed > approach should be regarded as any more promising than any other. Be patient. My hypotheses (which are not just spontaneous conjectures, but are based on an evaluation of the available evidence, the theoretical alternatives, and the logical and methodological problems involved) will be tested. They even have a potential connectionist component (in the induction of the features subserving categorization), although connectionism comes in for criticism too. For now it would seem only salutary to attempt to set cognitive modeling in directions that differ from the unprofitable ones it has taken so far. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 11 Jun 87 15:24:22 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <828@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., > Cambridge, MA writes: > > > There's no [symbol] grounding problem, just the old > > behavior-generating problem > > ..... There is: > (1) the behavior-generating problem (what I have referred to as the problem of > devising a candidate that will pass the Total Turing Test), (2) the > symbol-grounding problem (the problem of how to make formal symbols > intrinsically meaningful, independent of our interpretations), and (3) ... Just incidentally, what is the intrinsic meaning of "intrinsically meaningful"? The Turing test is an objectively verifiable criterion. How can we objectively verify intrinsic meaningfulness? > .... Add to this the surprising logical consequence that a > "dedicated" digital system (hardwired to its peripherals) would be > "analog" in its invertible inputs and outputs according to my > invertibility criterion, ..... Using "analog" to mean "invertible" invites misunderstanding, which invites irrelevant criticism. Human (in general, vertebrate) visual processing is a dedicated hardwired digital system. It employs data reduction to abstract such features as motion, edges, and orientation of edges. It then forms a map in which position is crudely analog to the visual plane, but quantized. This map is sufficiently similar to maps used in image processing machines so that I can almost imagine how symbols could be generated from it. By the time it gets to perception, it is not invertible, except with respect to what is perceived. Noninvertibility is demonstrated in experiments in the identification of suspects. Witnesses can report what they perceive, but they don't always perceive enough to invert the perceived image and identify the object that gave rise to the perception. If you don't agree, please give a concrete, objectively verifiable definition of "invertibility" that can be used to refute my conclusion. If I am right, human intelligence itself relies on neither analog nor invertible symbol grounding, and therefore artificial intelligence does not require it. By the way, there is an even simpler argument: even the best of us can engage in fuzzy thinking in which our symbols turn out not to be grounded. Subjectively, we then admit that our symbols are not intrinsically meaningful, though we had interpreted them as such. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Wed Jun 17 03:52:33 1987 Date: Wed, 17 Jun 87 03:52:21 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #147 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Wed, 17 Jun 87 03:17 EDT Received: from relay.cs.net by RELAY.CS.NET id af14824; 16 Jun 87 3:12 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10856; 16 Jun 87 3:10 EDT Date: Mon 15 Jun 1987 23:23-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #147 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 16 Jun 1987 Volume 5 : Issue 147 Today's Topics: Theory - Symbol Grounding and Physical Invertibility ---------------------------------------------------------------------- Date: 11 Jun 87 21:15:31 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In article <828@mind.UUCP> Stevan Harnad writes > >> There's no [symbol] grounding problem, just the old >> behavior-generating problem > There is: >(1) the behavior-generating problem (what I have referred to as the problem of >devising a candidate that will pass the Total Turing Test), (2) the >symbol-grounding problem (the problem of how to make formal symbols >intrinsically meaningful, independent of our interpretations), and (3) >the conjecture (based on the existing empirical evidence and on >logical and methodological considerations) that (2) is responsible for >the failure of the top-down symbolic approach to solve (1). It seems to me that in different places, you are arguing the relation between (1) and (2) in both directions, claiming both (A) The symbols in a purely symbolic system will always be ungrounded because such systems can't generate real performance; and (B) A purely symbolic system can't generate real performance because its symbols will always be ungrounded. That is, when I ask you why you think the symbolic approach won't work, one of your reasons is always "because it can't solve the grounding problem", but when I press you for why the symbolic approach can't solve the grounding problem, it always turns out to be "because I think it won't work." I think we should get straight on the priority here. It seems to me that, contra (3), thesis (A) is the one that makes perfect sense -- in fact, it's what I thought you were saying. I just don't understand (B) at all. To elaborate: I presume the "symbol-grounding" problem is a *philosophical* question: what gives formal symbols original intentionality? I suppose the only answer anybody knows is, in brief, that the symbols must be playing a certain role in what Dennett calls an "intentional system", that is, a system which is capable of producing complex, adaptive behavior in a rational way. Since such a system must be able to respond to changes in its environment, this answer has the interesting consequence that causal interaction with the world is a *necessary* condition for original intentionality. It tells us that symbols in a disconnected computer, without sense organs, could never be "grounded" or intrinsically meaningful. But those in a machine that can sense and react could be, provided the machine exhibited the requisite rationality. And this, as far as I can tell, is the end of what we learn from the "symbol grounding" problem -- you've got to have sense organs. For a system that is not causally isolated from the environment, the symbol-grounding problem now just reduces to the old behavior-generating problem, for, if we could just produce the behavior, there would be no question of the intentionality of the symbols. In other words, once we've wised up enough to recognize that we must include sensory systems (as symbolic AI has), we have completely disposed of the "symbol grounding" problem, and all that's left to worry about is the question of what kind of system can produce the requisite intelligent behavior. That is, all that's left is the old behavior-generating problem. Now as I've indicated, I think it's perfectly reasonable to suspect that the symbolic approach is insufficient to produce full human performance. You really don't have to issue any polemics on this point to me; such a suspicion could well be justified by pointing out the triviality of AI's performance achievements to date. What I *don't* see is any more "principled" or "logical" or "methodological" reason for such a suspicion; in particular, I don't understand how (B) could provide such a reason. My system can't produce intelligent performance because it doesn't make its symbols meaningful? This statement has just got things backwards -- if I could produce the behavior, you'd have to admit that its symbols had all the "grounding" they needed for original intentionality. In sum, apart from the considerations that require causal embedding, I don't see that there *is* any "symbol-grounding" problem, at least not any problem that is any different from the old "total-performance generating" problem. For this reason, I think your animadversions on symbol grounding are largely irrelevant to your position -- the really substantial claims pertain only to "what looks like it's likely to work" for generating intelligent behavior. On a more specific issue: > >You've bypassed the three points I brought up in replying to your >challenge to my invertibility criterion for an analog transform the >last time: (1) the quantization in standard A/D is noninvertible, Yes, but *my* point has been that since there isn't necessarily any more loss here than there is in a typical A/A transformation, the "degree of invertibility" criterion cross-cuts the intuitive A/D distinction. Look, suppose we had a digitized image, A, which is of much higher resolution than another analog one, B. A is more invertible since it contains more detail from which to reconstruct the original signal, but B is more "shape-preserving" in an intuitive sense. So, which do you regard as "more analog"? Which does your theory think is better suited to subserving our categorization performance? If you say B, then invertibility is just not what you're after. Anders Weinstein BBN Labs ------------------------------ Date: 12 Jun 87 08:16:00 EST From: cugini@icst-ecf.arpa Reply-to: Subject: symbol grounding and physical invertibility S. Harnad replies: > According to my view, invertibility (and perhaps inversion) > captures just the relevant features of causation and resemblance that > are needed to ground symbols. The relation is between the proximal > projection (of a distal object) onto the sensory surfaces -- let's > call it P -- and an invertible transformation of that projection [I(P)]. > The latter is what I call the "iconic representation." Note that the > invertibility is with the sensory projection, *not* the distal object. I > don't believe in distal magic. My grounding scheme begins at the > sensory surfaces ("skin and in"). No "wider" metaphysical causality is > involved, just narrow, local causality. Well, OK, glad you clarified that - I think there are issues here about the difference between grounding symbols in causation emanating from distal objects vs. grounding them in proximal sensory surfaces - (optical illusions, hallucinations, etc.) but let's pass over that for now. It still doesn't seem clear why invertibility should be necessary for grounding (although it may be sufficient). Frinstance, suppose we humans, or a robot, had four kinds of color receptors lurking behind our retinas (retinae?), which responded to red, green, blue and yellow wavelengths. And further suppose that stimulating the yellow receptors alone produced the same iconic representation as stimulating the red and green ones - ie both were experienced as plain old yellow, nor could the experiencer in any way distinguish between the yellows caused by the two different stimulations. (A fortiori, the experiencer would certainly not have more than one categorical representation, nor symbol for such experiences.) In short, suppose that some information was lost on the way in from the sensory surface, so we had a many to one (hence non-invertible) mapping. Would you then want to say that the symbol "yellow" was not grounded for such a being? John Cugini ------------------------------ Date: 12 Jun 87 15:52:40 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel writes: > Human visual processing is neither analog nor invertible. Nor understood nearly well enough to draw the former two conclusions, it seems to me. If you are taking the discreteness of neurons, the all-or-none nature of the action potential, and the transformation of stimulus intensity to firing frequency as your basis for concluding that visual processing is "digital," the basis is weak, and the analogy with electronic transduction strained. As the (unresolved) discussion of the logical basis of the A/D distinction last year indicated, nature itself may not be continuous, but quantized. This would make continuity-based definitions of A/D moot. If discrete photons strike discrete photoreceptors, then discontintuity is transforming into discontinuity. Yet the question can still be asked: Is the transformation preserving physical properties such as intensity and spatial relations by transforming them to physical properties that are isomorphic to them (e.g., intensity to frequency, and spatial adjacency to spatial adjacency) as opposed to merely "standing for" them in some binary code? There is also the question of postsynaptic potentials, which, unlike the all-or-none action potentials, are graded (to within the pharmacological quantum of a neurotransmitter packet). What if significant aspects of vision are coded at that level as fields or gradients and their interactions? Or at the level of local or distributed patterns of connectivity? Or at the chemical level? We don't even know how to match up the various resolution-levels or "grains" of the inputs and transformations involved: light quanta, neural quanta, psychophysical quanta. What is discrete and above-threshold at one level may become blurred, "continuous" and below-threshold at another. > what is the intrinsic meaning of "intrinsically meaningful"? > The Turing test is an objectively verifiable criterion. How can > we objectively verify intrinsic meaningfulness? We cannot objectively verify intrinsic meaningfulness. The Turing test is the only available criterion. Yet we can make inferences about it (for example, that it is unlikely to be present in a thermostat or lisp code running on a vax). And we have some direct (but subjective) evidence that it exists in at least one case (namely, our own): We know the difference between looking up a meaning in an English/English dictionary versus a Chinese/Chinese dictionary (if we are nonspeakers of Chinese): The former symbols are meaningful and the latter are not. We also know that we could "ground" an understanding of Chinese (by translation) in our prior understanding of English; and we assume that our understanding of English is grounded in our prior perceptual learning and understanding of categories in the real world of objects. Objective evidence of this perceptual grounding is provided by our ability to discriminate, manipulate, categorize, name and describe real-world objects and our ability to produce and respond to names and descriptions meaningfully (i.e., all Turing criteria). So the empirical question becomes the following: Is a device that has nothing but symbols and can only manipulate them on the basis of their shape more likely to be like our own (intrinsically grounded) case, or more like the Chinese/Chinese dictionary, whose meanings can only be derived by the mediation of an intrinsically grounded system like our own? But the issue is ultimately empirical. The logical and methodological considerations can really only serve to motivate pursuing one empirical hypothesis rather than another (e.g., top-down symbolic vs. bottom-up hybrid). The final arbiter is the Total Turing Test. If a pure symbolic module linked to transducers and effectors turns out to be able to generate all of our performance capacity then the grounding problem and intrinsic intentionality was a red herring. As I make clear in the paper "Minds, Machines and Searle," this is an empirical, not a logical question. But on the evidence to date, this outcome looks highly unlikely, and the obstacle seems to be the problem of bottom-up grounding of symbols in nonsymbolic representations and in the real world of objects. > Using "analog" to mean "invertible" invites misunderstanding, > which invites irrelevant criticism. I have tried to capture with the invertibility criterion certain features that may be important (perhaps even unique) to the case of cognitive modeling -- features that fail to be captured by the conventional electrical engineering criteria. I have acknowledged all along that the physically invertible/noninvertible distinction may turn out to be independent of the A/D distinction, although the overlap looks significant. And I'm doing my best to sort out the misunderstandings and irrelevant criticism... > Human (in general, vertebrate) visual processing is a dedicated > hardwired digital system. It employs data reduction to abstract such > features as motion, edges, and orientation of edges. It then forms a > map in which position is crudely analog to the visual plane, but > quantized. This map is sufficiently similar to maps used in image > processing machines so that I can almost imagine how symbols could be > generated from it. I am surprised that you state this with such confidence. In particular, do you really think that vertebrate vision is well enough understood functionally to draw such conclusions? And are you sure that the current hardware and signal-analytic concepts from electrical engineering are adequate to apply to what we do know of visual neurobiology, rather than being prima facie metaphors? > By the time it gets to perception, it is not invertible, except with > respect to what is perceived. Noninvertibility is demonstrated in > experiments in the identification of suspects. Witnesses can report > what they perceive, but they don't always perceive enough to invert > the perceived image and identify the object that gave rise to the > perception. If you don't agree, please give a concrete, objectively > verifiable definition of "invertibility" that can be used to refute my > conclusion. If I am right, human intelligence itself relies on neither > analog nor invertible symbol grounding, and therefore artificial > intelligence does not require it. I cannot follow your argument at all. Inability to categorize and identify is indeed evidence of a form of noninvertibility. But my theory never laid claim to complete invertibility throughout. (For the disadvantages of "total invertibility," see Luria's "The Mind of a Mnemonist," or, for a more literary depiction of the same problem, Borges's "Funes the Memorious." Both are discussed in a chapter of mine entitled "Metaphor and Mental Duality" in Simon & Sholes [eds] book "Language, Mind and Brain," Academic Press 1978. See also the literature on eidetic imagery.] Categorization and identification itself *requires* selective non-invertibility: within-category differences must be ignored and diminished, while between-category differences must be selected and enhanced. Although I do my best, it is not always possible to get all the relevant background material for these Net discussions onto the Net. Sometimes I must reluctantly refer discussants to a fuller text elsewhere. In principle, though, I'm prepared to re-present any particular piece of relevant material here. This particular misunderstanding, though, sounds like it would call for the exposition of my entire theory of categorization, which I am reluctant to impose on the entire Net without a wider demand. So let me just say that invertibility is my provisional criterion for what counts as an analog transformation, and that I have claimed that symbolic representations must be grounded in nonsymbolic ones, which include both invertible (iconic) and noninvertible (categorical) representations. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: Sun 14 Jun 87 16:42:34-PDT From: Ken Laws Reply-to: AIList-Request@STRIPE.SRI.COM Subject: [mind!harnad@princeton.edu (Stevan Harnad): Re: The symbol grounding problem] Date: 12 Jun 87 15:52:40 GMT From: mind!harnad@princeton.edu (Stevan Harnad) If discrete photons strike discrete photoreceptors, then discontintuity is transforming into discontinuity. Yet the question can still be asked: Is the transformation preserving physical properties such as intensity and spatial relations by transforming them to physical properties that are isomorphic to them (e.g., intensity to frequency, and spatial adjacency to spatial adjacency) as opposed to merely "standing for" them in some binary code? This makes me uncomfortable. Consider a "hash transformation" that maps a set of "intuitively meaningful" numeric symbols to a set of seemingly random binary codes. Suppose that the transformation can be computed by some [horrendous] information-preserving mapping of the reals to the reals. Now, the hash function satisfies my notion of an analog transformation (in the signal-processing sense). When applied to my discrete input set, however, the mapping does not seem to be analog (in the sense of preserving isomorphic relationships between pairs -- or higher orders -- of symbolic codes). Since information has not been lost, however, it should be possible to define "relational functions" that are analogous to "adjacency" and other properties in the original domain. Once this is done, surely the binary codes must be viewed as isomorphic to the original symbols rather than just "standing for them". The "information" in a signal is a function of your methods for extracting and interpreting the information. Likewise the "analog nature" of an information-preserving transformation is a function of your methods for decoding the analog relationships. We should also keep in mind that information theorists have advanced a great deal since the days of Shannon. Perhaps they have too limited (or general!) a view of information, but they have certainly considered your problem of decoding signal shape (as opposed to detecting modulation patterns). I regret that I am not familiar with their results, but I am sure that methods for decoding both discrete and continuous information in continuous signals are well studied. Not that all the answers are in -- vision workers like myself are well aware that there can be [obvious] information in a signal that is impossible to extract without a good model of the generating process. -- Ken ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Wed Jun 17 03:52:50 1987 Date: Wed, 17 Jun 87 03:52:35 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #148 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Wed, 17 Jun 87 03:20 EDT Received: from relay.cs.net by RELAY.CS.NET id ak14824; 16 Jun 87 3:14 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10883; 16 Jun 87 3:13 EDT Date: Mon 15 Jun 1987 23:37-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #148 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 16 Jun 1987 Volume 5 : Issue 148 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 15 Jun 87 13:23:35 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem (Reply to Ken Laws on ailist) Ken Laws on ailist@Stripe.SRI.Com writes: > Consider a "hash transformation" that maps a set of "intuitively > meaningful" numeric symbols to a set of seemingly random binary codes. > Suppose that the transformation can be computed by some [horrendous] > information-preserving mapping of the reals to the reals. Now, the > hash function satisfies my notion of an analog transformation (in the > signal-processing sense). When applied to my discrete input set, > however, the mapping does not seem to be analog (in the sense of > preserving isomorphic relationships between pairs -- or higher > orders -- of symbolic codes). Since information has not been lost, > however, it should be possible to define "relational functions" that > are analogous to "adjacency" and other properties in the original > domain. Once this is done, surely the binary codes must be viewed > as isomorphic to the original symbols rather than just "standing for > them". I don't think I disagree with this. Don't forget that I bit the bullet on some surprising consequences of taking my invertibility criterion for an analog transform seriously. As long as the requisite information-preserving mapping or "relational function" is in the head of the human interpreter, you do not have an invertible (hence analog) transformation. But as soon as the inverse function is wired in physically, producing a dedicated invertible transformation, you do have invertibility, even if a lot of the stuff in between is as discrete, digital and binary as it can be. I'm not unaware of this counterintuitive property of the invertibility criterion -- or even of the possibility that it may ultimately do it in as an attempt to capture the essential feature of an analog transform in general. Invertibility could fail to capture the standard A/D distinction, but may be important in the special case of mind-modeling. Or it could turn out not to be useful at all. (Although Ken Laws's point seems to strengthen rather than weaken my criterion, unless I've misunderstood.) Note, however, that what I've said about the grounding problem and the role of nonsymbolic representations (analog and categorical) would stand independently of my particular criterion for analog; substituting a more standard one leaves just about all of the argument intact. Some of the prior commentators (not Ken Laws) haven't noticed that, criticizing invertibility as a criterion for analog and thinking that they were criticizing the symbol grounding problem. > The "information" in a signal is a function of your methods for > extracting and interpreting the information. Likewise the "analog > nature" of an information-preserving transformation is a function > of your methods for decoding the analog relationships. I completely agree. But to get the requisite causality I'm looking for, the information must be interpretation-independent. Physical invertibility seems to give you that, even if it's generated by hardwiring the encryption/decryption (encoding/decoding) scheme underlying the interpretation into a dedicated system. > Perhaps [information theorists] have too limited (or general!) > a view of information, but they have certainly considered your > problem of decoding signal shape (as opposed to detecting modulation > patterns)... I am sure that methods for decoding both discrete and > continuous information in continuous signals are well studied. I would be interested to hear from those who are familiar with such work. It may be that some of it is relevant to cognitive and neural modeling and even the symbol grounding problems under discussion here. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 12 Jun 87 18:14:08 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., Cambridge, MA writes: > [1] [The only thing] we learn from the "symbol grounding" problem [is > that] you've got to have sense organs. > [2] For a system that is not causally isolated from the environment, > the symbol-grounding problem now just reduces to the old > behavior-generating problem, for, if we could just produce the behavior, > there would be no question of the intentionality of the symbols... > [3] [But claiming that a] system can't produce intelligent > performance *because* it doesn't make its symbols meaningful... has > just got things backwards -- if I could produce the behavior, you'd > have to admit that its symbols had all the "grounding" they needed > for original intentionality. > [4] For this reason, I think your animadversions on symbol > grounding are largely irrelevant to your position -- the really > substantial claims pertain only to "what looks like it's likely to > work" for generating intelligent behavior. [1] No, we don't merely learn that you need sense organs from the symbol grounding problem; we also learn that the nature of those sense organs, and their functional inter-relation with whatever else is going on downstream, may not be as simple as one might expect. The relation may be non-modular. It may not be just a matter of a simple hookup between autonomous systems -- sensory and symbolic -- as it is in current toy models. I agree that the symbol grounding problem does not logically entail this further conclusion, but it, together with the data, does suggest it, and why it might be important for generating successful performance. [2] I completely agree that a system that could pass the Total Turing Test using nothing but an autonomous symbolic module hooked to simple transducers would not be open to question about its "intrinsic intentionality" (at least not from groundedness considerations of the kind I've been describing here). But there's nothing circular about arguing that skepticism about the possibility of successfully passing the Total Turing Test with such a system is dictated in part by grounding considerations. The autonomy of the symbolic level can be the culprit in both respects. It can be responsible for the performance failures *and* for the lack of intrinsic intentionality. [3] Nor is there anything "backwards" about blaming the lack of intrinsic intentionality for performance failures. Rather, *you* may be engaging in counterfactual conditionals here. [4] The symbol grounding problem can hardly be irrelevant to my substantive hypotheses about what may work, since it is not only the motivation for them, but part of the explanation of why and how they may work. > since there isn't necessarily any more loss [in A/D] than there is > in a typical A/A transformation, the "degree of invertibility" > criterion cross-cuts the intuitive A/D distinction.... suppose we > had a digitized image, A, which is of much higher resolution > than another analog one, B. A is more invertible since it contains > more detail from which to reconstruct the original signal, but B is > more "shape-preserving" in an intuitive sense. So, which do you regard > as "more analog"? Which does your theory think is better suited to > subserving our categorization performance? If you say B, then > invertibility is just not what you're after. First, if A, the digital representation, is part of a dedicated system, hardwired to inputs and outputs, and the input stimuli are invertible, then, as I've said before, the whole system would be "analog" according to my provisional criterion, perhaps even more analog than B. If A is not part of a dedicated, physically invertible system, the question is moot, since it's not analog at all. With equal invertibility, it is an empirical question which is better suited to subserve cognition in general, and probably depends on optimality and capacity considerations. Finally, categorization performance in particular calls for much more than invertibility, as I've indicated before. Only iconic representations are invertible. Categorical reprsentations require selective *noninvertibility*. But that is another discussion... -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 12 Jun 87 21:36:13 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <6521@diamond.BBN.COM>, aweinste@Diamond.BBN.COM (Anders Weinstein) writes: > .... > (A) The symbols in a purely symbolic system will always be > ungrounded because such systems can't generate real performance; > ... > It seems to me that .... thesis (A) is the one that makes perfect > sense .... > > ..... I think it's perfectly reasonable to suspect that the > symbolic approach is insufficient to produce full human performance.... What exactly is this "purely" symbolic approach? What impure approach might be necessary? "Purely symbolic" sounds like a straw man: a system so purely abstract that it couldn't possibly relate to the real world, and that nobody seriously trying to mimic human behavior would even try to build anything that pure. To begin with, any attempt to "produce full human performance" must involve sensors, effectors, and motivations. Does "purely symbolic" preclude any of these? If not, what is it in the definition of a "purely symbolic" approach that makes it inadequate to pull these factors together? (Why do I so casually include motivations? I'm an amateur actor. Not even a human can mimic another human without knowing about motivations.) M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 12 Jun 87 22:19:48 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In article <837@mind.UUCP> Stevan Harnad (harnad@mind.UUCP) writes: > But there's nothing circular about >arguing that skepticism about the possibility of successfully passing >the Total Turing Test with such a system is dictated in part by >grounding considerations. The autonomy of the symbolic level can be >the culprit in both respects. It can be responsible for the performance >failures *and* for the lack of intrinsic intentionality. I'm afraid I still don't understand this. You write here as if these are somehow two *different* things. I don't see them that way, and hence find circularity. That is, I view intentionality as a matter of rational behavior. For me, the behavior is primary, and the notion of "symbol grounding" or "intrinsic intentionality" is conceptually derivative; and I thought from your postings that you shared this frankly behavioristic philosophy. Baldly put, here is the only plausible theory I know of "symbol grounding": X has intrinsic intentionality (is "grounded") iff X can pass the TTT. If you have a better theory, I'd like to hear it, but until then I believe that TTT-behavior is the very essence of intrinsic intentionality. Note that since it's the behavior that has conceptual priority, it makes sense to say that failure on the behavior front is, in a philosophical sense, the *reason* for a failure to make intrinsic intentionality. But to say the reverse is vacuous: failure to make intrinsic intentionality just *is the same thing* as failure to produce full TTT performance. I don't see that you can significantly distinguish these two failings. So what could it come to to say that symbolic AI must inevitably choke on the grounding problem? Since grounding == behavioral capability, all this claim can mean is that symbolic AI won't be able to generate full TTT performance. I think, incidentally, that you're probably right in this claim. However, I also think that the supposed "symbol-grounding problem" *is* irrelevant. From my point of view, it's just a fancy alternative name for the real issue, the behavior-generating problem. >[4] The symbol grounding problem can hardly be irrelevant to my >substantive hypotheses about what may work, since it is not only the >motivation for them, but part of the explanation of why and how they >may work. I still don't see how it explains anything. The grounding problem *reduces* to the behavior problem, not the other way around. To say that your approach is better grounded is only to say that it may work better (ie. generate TTT performance); there's just no independent content to the claim of "groundedness". Or do you have some non-behavioral definition of intrinsic intentionality that I haven't yet heard? Anders Weinstein BBN Labs ------------------------------ Date: 13 Jun 87 19:59:12 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., Cambridge, MA writes: > X has intrinsic intentionality (is "grounded") iff X can pass the TTT. > I thought from your postings that you shared this frankly behavioristic > philosophy... So what could it come to to say that symbolic AI must > inevitably choke on the grounding problem? Since grounding == behavioral > capability, all this claim can mean is that symbolic AI won't be able > to generate full TTT performance. I think, incidentally, that you're > probably right in this claim. However,...To say that your approach > is better grounded is only to say that it may work better (ie. > generate TTT performance); there's just no independent content to the > claim of "groundedness". Or do you have some non-behavioral definition > of intrinsic intentionality that I haven't yet heard? I think that this discussion has become repetitious, so I'm going to have to cut down on the words. Our disagreement is not substantive. I am not a behaviorist. I am a methodological epiphenomenalist. Intentionality and consciousness are not equivalent to behavioral capacity, but behavioral capacity is our only objective basis for inferring that they are present. Apart from behavioral considerations, there are also functional considerations: What kinds of internal processes (e.g., symbolic and nonsymbolic) look as if they might work? and why? and how? The grounding problem accordingly has functional aspects too. What are the right kinds of causal connections to ground a system? Yes, the test of successful grounding is the TTT, but that still leaves you with the problem of which kinds of connections are going to work. I've argued that top-down symbol systems hooked to transducers won't, and that certain hybrid bottom-up systems might. All these functional considerations concern how to ground symbols, they are distinct from (though ultimately, of course, dependent on) behavioral success, and they do have independent content. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 14 Jun 87 19:45:33 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In article <1163@houdi.UUCP> marty1@houdi.UUCP (M.BRILLIANT) writes: >> (A) The symbols in a purely symbolic system ... > >What exactly is this "purely" symbolic approach? What impure approach >might be necessary? "Purely symbolic" sounds like a straw man ... The phrase "purely symbolic" was just my short label for the AI strategy that Stevan Harnad has been criticizing. Yes this strategy *does* encompass the use of sensors and effectors and (maybe) motivations. Sorry if the term was misleading, I was only using it as pointer; consult Harnad's postings for a fuller characterization. ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Wed Jun 17 03:53:11 1987 Date: Wed, 17 Jun 87 03:52:51 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #149 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Wed, 17 Jun 87 03:23 EDT Received: from relay.cs.net by RELAY.CS.NET id ae14895; 16 Jun 87 3:26 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10935; 16 Jun 87 3:25 EDT Date: Mon 15 Jun 1987 23:42-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #149 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 16 Jun 1987 Volume 5 : Issue 149 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 14 Jun 87 15:13:34 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <843@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > > Intentionality and consciousness are not equivalent to behavioral > capacity, but behavioral capacity is our only objective basis for > inferring that they are present. Apart from behavioral considerations, > there are also functional considerations: What kinds of internal > processes (e.g., symbolic and nonsymbolic) look as if they might work? > and why? and how? The grounding problem accordingly has functional aspects > too. What are the right kinds of causal connections to ground a > system? Yes, the test of successful grounding is the TTT, but that > still leaves you with the problem of which kinds of connections are > going to work. I've argued that top-down symbol systems hooked to > transducers won't, and that certain hybrid bottom-up systems might. All > these functional considerations concern how to ground symbols, they are > distinct from (though ultimately, of course, dependent on) behavioral > success, and they do have independent content. Harnad's terminology has proved unreliable: analog doesn't mean analog, invertible doesn't mean invertible, and so on. Maybe top-down doesn't mean top-down either. Suppose we create a visual transducer feeding into an image processing module that could delineate edges, detect motion, abstract shape, etc. This processor is to be built with a hard-wired capability to detect "objects" without necessarily finding symbols for them. Next let's create a symbol bank, consisting of a large storage area that can be partitioned into spaces for strings of alphanumeric characters, with associated pointers, frames, anything else you think will work to support a sophisiticated knowledge base. The finite area means that memory will be limited, but human memory can't really be infinite, either. Next let's connect the two: any time the image processor finds an object, the machine makes up a symbol for it. When it finds another object, it makes up another symbol and links that symbol to the symbols for any other objects that are related to it in ways that it knows about (some of which might be hard-wired primitives): proximity in time or space, similar shape, etc. It also has to make up symbols for the relations it relies on to link objects. I'm over my head here, but I don't think I'm asking for anything we think is impossible. Basically, I'm looking for an expert system that learns. Now we decide whether we want to play a game, which is to make the machine seem human, or whether we want the machine to exhibit human behavior on the same basis as humans, that is, to survive. For the game, the essential step is to make the machine communicate with us both visually and verbally, so it can translate the character strings it made up into English, so we can understand it and it can understand us. For the survival motivation, the machine needs a full set of receptors and effectors, and an environment in which it can either survive or perish, and if we built it right it will learn English for its own reasons. It could also endanger our survival. Now, Harnad, Weinstein, anyone: do you think this could work, or do you think it could not work? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 14 Jun 87 14:15:55 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <835@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel writes: > > > Human visual processing is neither analog nor invertible. > > Nor understood nearly well enough to draw the former two conclusions, > it seems to me. If you are taking the discreteness of neurons, the > all-or-none nature of the action potential, and the transformation of > stimulus intensity to firing frequency as your basis for concluding > that visual processing is "digital," the basis is weak, and the > analogy with electronic transduction strained. No, I'm taking more than that as the basis. I don't have any names handy, and I'm not a professional in neurobiology, but I've seen many articles in Science and Scientific American (including a classic paper titled something like "What the frog's eye tells the frog's brain") that describe the flow of visual information through the layers of the retina, and through the layers of the visual cortex, with motion detection, edge detection, orientation detection, etc., all going on in specific neurons. Maybe a neurobiologist can give a good account of what all that means, so we can guess whether computer image processing could emulate it. > > what is the intrinsic meaning of "intrinsically meaningful"? > > The Turing test is an objectively verifiable criterion. How can > > we objectively verify intrinsic meaningfulness? > > We cannot objectively verify intrinsic meaningfulness. The Turing test > is the only available criterion. Yet we can make inferences... I think that substantiates Weinstein's position: we're back to the behavior-generating problem. > ....: We > know the difference between looking up a meaning in an English/English > dictionary versus a Chinese/Chinese dictionary (if we are nonspeakers > of Chinese): The former symbols are meaningful and the latter are > not. Not relevant. Intrinsically, words in both languages are equally meaningful. > > Using "analog" to mean "invertible" invites misunderstanding, > > which invites irrelevant criticism. > > ..... I have acknowledged all > along that the physically invertible/noninvertible distinction may > turn out to be independent of the A/D distinction, although the > overlap looks significant. And I'm doing my best to sort out the > misunderstandings and irrelevant criticism... Then please stop using the terms analog and digital. > > > Human (in general, vertebrate) visual processing is a dedicated > > hardwired digital system. It employs data reduction to abstract such > > features as motion, edges, and orientation of edges. It then forms a > > map in which position is crudely analog to the visual plane, but > > quantized. This map is sufficiently similar to maps used in image > > processing machines so that I can almost imagine how symbols could be > > generated from it. > > I am surprised that you state this with such confidence. In > particular, do you really think that vertebrate vision is well enough > understood functionally to draw such conclusions? ... Yes. See above. > ... And are you sure > that the current hardware and signal-analytic concepts from electrical > engineering are adequate to apply to what we do know of visual > neurobiology, rather than being prima facie metaphors? Not the hardware concepts. But I think some principles of information theory are independent of the medium. > > By the time it gets to perception, it is not invertible, except with > > respect to what is perceived. Noninvertibility is demonstrated in > > experiments in the identification of suspects. Witnesses can report > > what they perceive, but they don't always perceive enough to invert > > the perceived image and identify the object that gave rise to the > > perception.... > > .... If I am right, human intelligence itself relies on neither > > analog nor invertible symbol grounding, and therefore artificial > > intelligence does not require it. > > I cannot follow your argument at all. Inability to categorize and identify > is indeed evidence of a form of noninvertibility. But my theory never laid > claim to complete invertibility throughout..... First "analog" doesn't mean analog, and now "invertibility" doesn't mean complete invertibility. These arguments are getting too slippery for me. > .... Categorization and identification > itself *requires* selective non-invertibility: within-category differences > must be ignored and diminished, while between-category differences must > be selected and enhanced. Well, that's the point I've been making. If non-invertibility is essential to the way we process information, you can't say non-invertibility would prevent a machine from emulating us. Anybody can do hand-waving. To be convincing, abstract reasoning must be rigidly self-consistent. Harnad's is not. I haven't made any assertions as to what is possible. All I'm saying is that Harnad has come nowhere near proving his assertions, or even making clear what his assertions are. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 15 Jun 87 01:43:55 GMT From: berleant@sally.utexas.edu (Dan Berleant) Subject: Re: The symbol grounding problem It is interesting that some (presumably significant) visual processing occurs by graded potentials without action potentials. Receptor cells (rods & cones), 'horizontal cells' which process the graded output of the receptors, and 'bipolar cells' which do further processing, use no action potentials to do it. This seems to indicate the significance of analog processing to vision. There may also be significant invertibility at these early stages of visual processing in the retina: One photon can cause several hundred sodium channels in a rod cell to close. Such sensitivity suggests a need for precise representation of visual stimuli which suggests the representation might be invertible. Furthermore, the retina cannot be viewed as a module, only loosely coupled to the brain. The optic nerve, which does the coupling, has a high bandwidth and thus carries much information simultaneously along many fibers. In fact, the optic nerve carries a topographic representation of the retina. To the degree that a topographic representation is an iconic representation, the brain thus receives an iconic representation of the visual field. Furthermore, even central processing of visual information is characterized by topographic representations. This suggests that iconic representations are important to the later stages of perceptual processing. Indeed, all of the sensory systems seem to rely on topographic representations (particularly touch and hearing as well as vision). An interesting example in hearing is direction perception. Direction seems to be, as I understand it, found by processing the difference in time from when a sound reaches one ear to when it reaches the other, in large part. The resulting direction is presumably an invertible representation of that time difference. Dan Berleant UUCP: {gatech,ucbvax,ihnp4,seismo,kpno,ctvax}!ut-sally!berleant ARPA: ai.berleant@r20.utexas.edu ------------------------------ Date: 14 Jun 87 15:03:51 GMT From: harwood@cvl.umd.edu (David Harwood) Subject: Re: The symbol grounding problem In article <843@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes: > (... replying to Anders Weinstein ...who wonders "Where's the beef?" in Steve Harnad's conceptual and terminological salad ...; uh - let me be first to prophylactically remind us - lest there is any confusion and forfending that he should perforce of intellectual scruple must need refer to his modest accomplishments - Steve Harnad is editor of Behavioral and Brain Sciences, and I am not, of course. We - all of us - enjoy reading such high-class stuff...;-) Anyway, Steve Harnad replies to A.W., re "Total Turing Tests", behavior, and the (great AI) "symbol grounding problem": >I think that this discussion has become repetitious, so I'm going to >have to cut down on the words. Praise the Lord - some insight - by itself, worthy of Pass of the "Total Turing Test." >... Our disagreement is not substantive. >I am not a behaviorist. I am a methodological epiphenomenalist. I'm not a behaviorist, you're not a behaviorist, he's not a behaviorist too ... We are all methodological solipsists hereabouts on this planet, having already, incorrigibly, failed the "Total Turing Test" for genuine intergalactic First Class rational beings, but so what? (Please, Steve - this is a NOT a test - I repeat - this is NOT a test of your philosophical intelligence. It is an ACTUAL ALERT of your common sense, not to mention, sense of humor. Please do not solicit BBS review of this thesis... >... Apart from behavioral considerations, >there are also functional considerations: What kinds of internal >processes (e.g., symbolic and nonsymbolic) look as if they might work? >and why? and how? The grounding problem accordingly has functional aspects >too. What are the right kinds of causal connections to ground a >system? Yes, the test of successful grounding is the TTT, but that >still leaves you with the problem of which kinds of connections are >going to work. I've argued that top-down symbol systems hooked to >transducers won't, and that certain hybrid bottom-up systems might. All >these functional considerations concern how to ground symbols, they are >distinct from (though ultimately, of course, dependent on) behavioral >success, and they do have independent content. >-- > >Stevan Harnad (609) - 921 7771 >{bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad >harnad%mind@princeton.csnet harnad@mind.Princeton.EDU You know what is the real problem with your postings - it's what I would call "the symbol grounding problem". You want to say obvious things in the worst possible way, otherwise say abstract things in the worst possible way.. And ignore what others say. Also, for purposes of controversial public discussion, ignore scientific 'facts' (eg about neurologic perceptual equivalence), and standard usage of scientific terminology and interpretation of theories. (Not that these are sacrosanct.) It seems to me that your particular "symbol grounding problem" is indeed the the sine qua non of the Total Turing Test for "real" philosophers of human cognition. As I said, we are all methodological solipsists hereabouts. However, if you want AI funding from me - I want to see what real computing system, using your own architecture and object code of at least 1 megabytes, has been designed by you. Then we will see how your "symbols" are actually grounded, using the standard, naive but effective denotational semantics for the "symbols" of your intention, qua "methodological epiphenomensist." David Harwood ------------------------------ Date: 15 Jun 87 03:32:07 GMT From: berleant@sally.utexas.edu (Dan Berleant) Subject: Re: The symbol grounding problem In article <835@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes: >We cannot objectively verify intrinsic meaningfulness. The Turing test >is the only available criterion. Yes, the Turing test is by definition subjective, and also subject to variable results from hour to hour even from the same judge. But I think I disagree that intrinsic meaningfulness cannot be objectively verified. What about the model theory of logic? Dan Berleant UUCP: {gatech,ucbvax,ihnp4,seismo,kpno,ctvax}!ut-sally!berleant ARPA: ai.berleant@r20.utexas.edu ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Wed Jun 17 03:53:26 1987 Date: Wed, 17 Jun 87 03:53:04 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #150 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Wed, 17 Jun 87 03:25 EDT Received: from relay.cs.net by RELAY.CS.NET id ab14916; 16 Jun 87 3:32 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10975; 16 Jun 87 3:29 EDT Date: Mon 15 Jun 1987 23:45-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #150 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 16 Jun 1987 Volume 5 : Issue 150 Today's Topics: Binding - comp.theory Newsgroup, Theory - Information Flow & The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 11 Jun 87 21:17:29 GMT From: ramesh@cvl.umd.edu (Ramesh Sitaraman) Subject: New newsgroup: comp.theory To those of you haven't already noticed .... A new news group "comp.theory" has commenced. This group presumably deals with all aspects of the Theoretcial Computer Science including Complexity theory, Algorithm analysis, Logic and theory of computation, denotational semantics, computational geometry etc etc etc. Make merry, Ramesh ------------------------------ Date: Wed 10 Jun 87 12:31:39-EDT From: Albert Boulanger Subject: Re: Information flow discussions Anthony Pelletier writes: P.S. I think alot about information flow problems and would enjoy discussions on that...if anyone wants to chat. For a real "juicy" discussion of information flow in non-linear systems see: "Strange Attractors, Chaotic Behavior, and Information Flow" Robert Shaw, Z. Naturforsch, 36a, 80-112 1981 This discusses the information flow characteristics of non-linear systems in order to gain insight on how non-linear systems self-organize. (This self-organization aspect of non-linear dynamical systems is an aspect of neural networks. See for example Kohonen's work on self-organizing feature maps in "Self-Organization and Associative Memory" Springer-Verlag 1984. This feature map stuff is a type of unsupervised learning.) Albert Boulanger BBN Labs ------------------------------ Date: 15 Jun 87 02:37:00 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem In two consecutive postings marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel wrote: > the flow of visual information through the layers of the retina, > and through the layers of the visual cortex, with motion detection, > edge detection, orientation detection, etc., all going on in specific > neurons... Maybe a neurobiologist can give a good account of what > all that means, so we can guess whether computer image > processing could emulate it. As I indicated the last time, neurobiologists don't *know* what all those findings mean. It is not known how features are detected and by what. The idea that single cells are doing the detecting is just a theory fragment, and one that has currently fallen on hard times. Rivals include distributed networks (of which the cell is just a component), or spatial frequency detectors, or coding at some entirely different level, such as continuous postsynaptic potentials, local circuits, architectonic columns or neurochemistry. Some even think that the multiple analog retinas at various levels of the visual system (12 on each side, at last count) may have something to do with feature extraction. One cannot just take current neurophysiological data and replace the nonexistent theory by preconceptions from machine vision -- especially not by way of justifying the machine-theoretic concepts. >> >[SH:] my theory never laid claim to complete invertibility >> >throughout. > > First "analog" doesn't mean analog, and now "invertibility" > doesn't mean complete invertibility. These arguments are > getting too slippery for me... If non-invertibility is essential > to the way we process information, you can't say non-invertibility > would prevent a machine from emulating us. I have no idea what proposition you think you were debating here. I had pointed out a problem with the top-down symbolic approach to mind-modeling -- the symbol grounding problem -- which suggested that symbolic representations would have to be grounded in nonsymbolic representations. I had also sketched a model for categorization that attempted to ground symbolic representations in two nonsymbolic kinds of representations -- iconic (analog) representations and categorical (feature-filtered) representations. I also proposed a criterion for analog transformations -- invertibility. I never said that categorical representations were invertible or that iconic representations were the only nonsymbolic representations you needed to ground symbols. Indeed, most of the CP book under discussion concerns categorical representations. > All I'm saying is that Harnad has come nowhere near proving his > assertions, or even making clear what his assertions are... > Harnad's terminology has proved unreliable: analog doesn't mean > analog, invertible doesn't mean invertible, and so on. Maybe > top-down doesn't mean top-down either... > Anybody can do hand-waving. To be convincing, abstract > reasoning must be rigidly self-consistent. Harnad's is not. > I haven't made any assertions as to what is possible. Invertibility is my candidate criterion for an analog transform. Invertible means invertible, top-down means top-down. Where further clarification is needed, all one need do is ask. Now here is M. B. Brilliant's "Recipe for a symbol-grounder" (not to be confused with an assertion as to what is possible): > Suppose we create a visual transducer... with hard-wired > capability to detect "objects"... Next let's create a symbol bank > Next let's connect the two... I'm over my head here, but I don't > think I'm asking for anything we think is impossible. Basically, > I'm looking for an expert system that learns... the essential step > is to make the machine communicate with us both visually and verbally, > so it can translate the character strings it made up into English, so > we can understand it and it can understand us. For the survival > motivation, the machine needs a full set of receptors and > effectors, and an environment in which it can either survive or > perish, and if we built it right it will learn English for its > own reasons. Now, Harnad, Weinstein, anyone: do you think this > could work, or do you think it could not work? Sounds like a conjecture about a system that would pass the TTT. Unfortunately, the rest seems far too vague and hypothetical to respond to. If you want me to pay attention to further postings of yours, stay temperate and respectful as I endeavor to do. Dismissive rhetoric will not convince anyone, and will not elicit substantive discussion. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 15 Jun 87 05:21:36 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem berleant@ut-sally.UUCP (Dan Berleant) of U. Texas CS Dept., Austin, Texas has posted this welcome reminder: > the retina cannot be viewed as a module, only loosely > coupled to the brain. The optic nerve, which does the coupling, has a > high bandwidth and thus carries much information simultaneously along > many fibers. In fact, the optic nerve carries a topographic > representation of the retina. To the degree that a topographic > representation is an iconic representation, the brain thus receives an > iconic representation of the visual field. > Furthermore, even central processing of visual information is > characterized by topographic representations. This suggests that iconic > representations are important to the later stages of perceptual > processing. Indeed, all of the sensory systems seem to rely on > topographic representations (particularly touch and hearing as well as > vision). As I mentioned in my last posting, at last count there were 12 pairs of successively higher analog retinas in the visual system. No one yet knows what function they perform, but they certainly suggest that it is premature to dismiss the importance of analog representations in at least one well optimized system... > Yes, the Turing test is by definition subjective, and also subject to > variable results from hour to hour even from the same judge. > But I think I disagree that intrinsic meaningfulness cannot be > objectively verified. What about the model theory of logic? In earlier postings I distinguished between two components of the Turing Test. One is the formal, objective one: Getting a system to generate all of our behavioral capacities. The second is the informal, intuitive (and hence subjective) one: Can a person tell such a device apart from a person? This version must be open-ended, and is no better or worse than -- in fact, I argue that is identical to -- the real-life turing-testing we do of one another in contending with the "other minds" problem. The subjective verification of intrinsic meaning, however, is not done by means of the informal turing test. It is done from the first-person point of view. Each of us knows that his symbols (his linguistic ones, at any rate) are grounded, and refer to objects, rather than being menaningless syntactic objects manipulated on the basis of their shapes. I am not a model theorist, so the following reply may be inadequate, but it seems to me that the semantic model for an uninterpreted formal system in formal model-theoretic semantics is always yet another formal object, only its symbols are of a different type from the symbols of the system that is being interpreted. That seems true of *formal* models. Of course, there are informal models, in which the intended interpretation of a formal system corresponds to conceptual or even physical objects. We can say that the intended interpretation of the primitive symbol tokens and the axioms of formal number theory are "numbers," by which we mean either our intuitive concept of numbers or whatever invariant physical property quantities of objects share. But such informal interpretations are not what formal model theory trades in. As far as I can tell, formal models are not intrinsically grounded, but depend on our concepts and our linking them to real objects. And of course the intrinsic grounding of our concepts and our references to objects is what we are attempting to capture in confronting the symbol grounding problem. I hope model theorists will correct me if I'm wrong. But even if the model-theoretic interpretation of some formal symbol systems can truly be regarded as the "objects" to which it refers, it is not clear that this can be generalized to natural language or to the "language of thought," which must, after all, have Total-Turing-Test scope, rather than the scope of the circumscribed artificial languages of logic and mathematics. Is there any indication that all that can be formalized model-theoretically? -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Fri Jun 19 03:10:42 1987 Date: Fri, 19 Jun 87 03:10:33 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #151 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Fri, 19 Jun 87 03:06 EDT Received: from relay.cs.net by RELAY.CS.NET id ac02471; 18 Jun 87 15:42 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa10652; 18 Jun 87 15:41 EDT Date: Wed 17 Jun 1987 23:58-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #151 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Thursday, 18 Jun 1987 Volume 5 : Issue 151 Today's Topics: Seminars - Mechanization of Programmer's Knowledge (MCC) & Partial Order Programming (MCC) & AI at Vanderbilt & Comparative Induction (NASA Ames) & Default Reasoning and Stereotypes in User Modelling (UPenn), Conference - Last Call for AAAI-87 Volunteers & CADE-9: Automated Deduction & Office Knowledge & Second Eurographics Workshop on Intelligent CAD Systems ---------------------------------------------------------------------- Date: Mon 15 Jun 87 14:12:25-CDT From: Ellie Huck Subject: Seminar - Mechanization of Programmer's Knowledge (MCC) Please join the AI group for the following speaker: ON THE MECHANIZATION OF PROGRAMMER'S KNOWLEDGE Henryk Jan Komorowski Harvard University June 17 - 10:00 MCC Balcones Auditorium What do programmers with experience in writing programs know and how can this knowledge be mechanized so it can be used by a computer? This talk presents an informal overview of the foundations of mechanical support for software design. The goal of the mechanization is to provide an intelligent assistant for the programmer that can uncover flaws in the design rather than automatically generate programs. What programmer knows is divided into knowledge of data structures, recursive schemata, assimilation rules, and the process of designing a program which is similar to extension of a theory. A prototype system now implemented provides salient advice, despite its limited knowledge-base. June 17 - 10:00 MCC Balcones Auditorium ------------------------------ Date: Mon 15 Jun 87 14:50:06-CDT From: Ellie Huck Subject: Seminar - Partial Order Programming (MCC) Please join the AI Group for the following speaker: PARTIAL ORDER PROGRAMMING D. Stott Parker UCLA Computer Science Department June 19 - 10:00 MCC Balcones Auditorium We introduce a declarative programming paradigm that describes computation with partial orders. A partial order program corresponds to a collection of constraints u >= C(u) where >= is a partial order on a domain of `objects' and `values', u is an object, and C(u) is an object or a value. Semantics of such a program consist of assignments of values to the objects u that satisfy the inequalities. When C is a monotone and continuous function, fixpoint semantics of the program may be obtained easily and naturally. The partial order programming paradigm has interesting properties: (1) It generalizes various computational paradigms (logic, functional, object-oriented, and others) in a clean way. (2) It takes thorough advantage of known results for continuous functionals on partial orders, providing a clear semantics for the paradigm. (3) It presents a framework that may be more generally acceptable for dealing with `cognitive' computation problems, including natural language processing and knowledge representation. (4) It coincides with recent work on relaxation solution of a variety of problems including consistent labelling, path problems, and linear algebraic systems. June 19 - 10:00 MCC Balcones Auditorium ------------------------------ Date: Tue, 16 Jun 87 15:15:50 PDT From: JARED%PLU@ames-io.ARPA Subject: Seminars - AI at Vanderbilt & Comparative Induction (NASA Ames) NASA, Ames Research Center Intelligent Systems Forum TWO SPEAKERS: Dr. Arthur J. Brodersen Center for Intelligent Systems Vanderbilt University Expert Systems Research at Vanderbilt University Abstract: The current research activities at the Center for Intelligent Systems at Vanderbilt University will be discussed. The current research activities include knowledge-based systems for test technologies, intelligent tutorial systems for simulation tools, training systems, knowledge retrieval systems, and diagnositic and repair systems. David Hartzband Digital Equipment Corporation Chief Scientist, Artificial Intelligence Technology Group and Visiting Scholar, Stanford University Comparative Induction Methods for Problem Solving and (Some) Learning} Date: Thursday, June 18, 1987 Time: 2:00 to 4:30 PM Location: Bldg. 245, Space Science Auditorium Inquires: David Jared, (415) 694-6525, jared@ames-pluto.arpa VISITORS ARE WELCOME: Register and obtain vehicle pass at Ames Visitor Reception Building (N-253) or the Security Station near Gate 18. Do not use the Navy Main Gate. Non-citizens (except Permanent Residents) must have prior approval from the Director's Office one week in advance. Submit requests to the point of contact indicated above. Non-citizens must register at the Visitor Reception Building. Permanent Residents are required to show Alien Registration Card at the time of registration. ------------------------------ Date: 15 Jun 87 12:15:25 EDT From: Marcella.Zaragoza@isl1.ri.cmu.edu Subject: Seminar - Default Reasoning and Stereotypes in User Modelling (UPenn) SPECIAL SEMINAR SPEAKER: Timothy Finin Computer and Information Science University of Pennsylvania, Philadelphia, PA WHEN: Thursday, June 18, 1987, 10:00 am WHERE: Doherty Hall 3313 TOPIC: DEFAULT REASONING AND STEREOTYPES IN USER MODELLING This talk discusses the application of various kinds of default reasoning in systems which must maintain a model of its users. In particular, we describe a general architecture of a domain independent system for building and maintaining long term models of individual users. The user modelling system is intended to provide a well defined set of services for an application system which is interacting with various users and has a need to build and maintain models of them. As the application system interacts with a user, it can acquire knowledge of him and pass that knowledge on to the user model maintenance system for incorporation. We describe a prototype general user modelling system (hereafter called GUMS1 which we have implemented in Prolog. This system satisfies some of the desirable characteristics we discuss. ------------------------------ Date: 16 Jun 87 20:18:07 GMT From: feifer@locus.ucla.edu Subject: Conference - Last call for AAAI-87 volunteers Due to some last minute cancellations we have a few openings for volunteers for AAAI-87. Please see the original posting below for more information. If interested, please respond immediately. ANNOUNCEMENT: Student Volunteers Needed for Artificial Intelligence Conference AAAI-87 AAAI-87 (American Association on Artificial Intelligence) will be held July 13-17, 1987 in beautiful Seattle, Washington. Student volunteers are needed to help with local arrangements and staffing of the conference. To be eligible for a Volunteer position, an individual must be an undergraduate or graduate student in any field at any college or university. This is an excellent opportunity for students to participate in the conference. Volunteers receive FREE registration at AAAI-87, conference proceedings, "STAFF" T-shirt, and are invited to the volunteer party. More importantly, by participating as a volunteer, you become more involved and meet students and researchers with similar interests. Volunteer responsibilities are varied, including conference preparation, registration, staffing of sessions and tutorials and organizational tasks. Each volunteer will be assigned twelve (12) hours. If you are interested in participating in AAAI-87 as a Student Volunteer, apply by sending the following information: Name Electronic Mail Address USMail Address Telephone Number(s) Dates Available Student Affiliation Advisor's Name to: feifer@locus.ucla.edu or Richard Feifer UCLA Center for the Study of Evaluation 145 Moore Hall Los Angeles, California 90024 Thanks, and I hope you join us this year! Richard Feifer Student Volunteer Coordinator AAAI-87 Staff - Richard ------------------------------ Date: Mon, 15 Jun 87 20:07:49 cdt From: stevens@anl-mcs.ARPA (Rick L. Stevens) Subject: Conference - CADE-9: Automated Deduction Preliminary Announcement and Call for Papers 9th International Conference on Automated Deduction May 23-26, 1988 CADE-9 will be held at Argonne National Laboratory (near Chicago) in celebration of the 25th anniversary of the discovery of the resolution principle at Argonne in the sum- mer of 1963. Papers are invited in the following or related fields: Theorem Proving Logic Programming Unification Deductive Databases Term Rewriting ATP for Non-Standard Logics Program Verification Inference Systems The Program Committee includes: Peter Andrews Hans-Jorgen Ohlbach W.W. Bledsoe Ross Overbeek Alan Bundy William Pase Seif Haridi Jorg Siekmann Larry Henschen Jim Williams Jean-Louis Laissez Mark Stickel Dallas Lankford Ewing Lusk Michael MacRobbie Papers should be sent to arrive before November 23rd, 1987 to Ewing Lusk and Ross Overbeek, chairmen CADE-9 Mathematics and Computer Science Division Argonne National Laboratory Argonne, IL 60439 ------------------------------ Date: Mon, 15 Jun 87 17:17:53 edt From: rba@flash.bellcore.com (Robert B. Allen) Subject: Conference - Office Knowledge Office Knowledge: Representation, Management and Utilization University of Toronto IFIP WG8.4 Workshop Program For information contact: Fred Lochovsky, fred@csri.toronto.edu Monday, August 17th, 1987 8:00-9:00 Registration 9:00-9:15 Workshop Opening Remarks 9:15-10:45 Session: Invited Talk Objects and Things. D. Tsichritzis, Universite de Geneve, Switzerland 11:15-12:45 Session: Supporting Organizational Activities Ubik: A System for Conceptual and Organizational Development. P. de Jong, MIT, U.S.A. KNOOM - KNowledge Oriented Office Model Representation of Knowledge in the Office. M. Hofmann, Universitaet Wien, Austria OTM: A Language for Representing Concurrent Office Tasks. J. Hogg, University of Toronto, Canada 2:00-3:30 Session: Invited Talk Representing Office Work with Goals and Constraints. W.B. Croft, University of Massachusetts, U.S.A. 4:00-5:30 Session: Representing, Querying and Generating Office Objects Time Management in the Office-net System. R. Maiocchi, R. Zicari, Politecnico di Milano, Italy, M. Fugini, Universita di Brescia, Italy Towards a Graphic Query Interface for Complex Objects. G. Lausen, Universitaet Mannheim, West Germany, A. Oberweis, Technische Hochschule Darmstadt, West Germany Knowledge Representation and Utilization in Automatic Office Form Generation. K. Watabe, K. Tsuruoka, NEC Corporation, Japan 5:30-6:30 Reception 6:30-7:30 Demonstration Meta-Data for Automating the Management of Office Information. R.E.A. Mason, A. Benjamin, J.R. Tessier, Online People Inc., Canada _________________________________________________________________ Tuesday, August 18th, 1987 9:00-10:30 Session: Invited Talk Organizational Semantics. C. Hewitt, MIT, U.S.A. 11:00-12:30 Session: Problem Solving An Office Environment to Support Problem Solving. P.W.G. Bots, H.G. Sol, Delft University of Technology, Netherlands Generic Knowledge in Office Activities. A.A. Araya, M.J. Stefik, Xerox PARC, U.S.A. EXPERTNET: An Approach to Resource Sharing on a Network of Workstations. A. Allam, Northern Telecom Canada Ltd., Canada, G.M. White, University of Ottawa, Canada 2:00-3:30 Session: Text and Pictures Semantics and Conceptual Modelling of Documents. F. Barbic, S. Daneluzzi, F. Garzotto, S. Mainetti, P. Paolini, Politecnico di Milano, Italy Knowledge-Based Text Processing in Office Environments: The Text Condensation System TOPIC. U. Hahn, Universitaet Passau, West Germany, U. Reimer, Universitaet Konstanz, West Germany Knowledge Base for Storage and Retrieval of Pictures. B. Beetz, SEL Research Center, West Germany 4:00-5:30 Session: Poster Session Artificial Intelligence and Organizational Design: Prospects of Integrating Two Perspectives. U. Frank, Universitaet Mannheim, West Germany Intermediate Knowledge Representation for Extended Office Systems. E.S. Cordingley, University of Surrey, England Intelligent Interfaces for Office Information Systems. B.C. Desai, Concordia University, Canada, C. Frasson, J. Vaucher, Universite de Montreal, Canada Managing Office Knowledge through Conceptual Structures. G. Berg-Cross, Advanced Decision Systems, U.S.A. Picture Management on Optical Disks: A Practical Approach on Micro-computers. S. Miranda, N. Le Thanh, A.C. Salgado, E. Borelli-Vittori, Universite de Nice, France Managing Replicas in Distributed Telephone/Address Directories. H.M. Gladney, IBM Almaden Research Center, U.S.A. 7:30 Banquet _________________________________________________________________ Wednesday, August 19th, 1987 9:00-10:30 Session: Invited Talk NICK: Intelligent Computer Supported Cooperative Work. C. Ellis, MCC, U.S.A. 11:00-12:30 Session: Office Communication Solving the Connection Problem. M.S. Mazer, University of Toronto, Canada Viewing Communication as a Problem Solving Activity: An Enrichment Towards Supporting Cooperative Office Work. C.C. Woo, F.H. Lochovsky, University of Toronto, Canada CHAOS: A Knowledge-Based System for Conversing Inside Offices. F. De Cindio, C. Simone, R. Vassallo, A. Zanaboni, Universita di Milano, Italy 2:00-3:30 IFIP WG8.4 Business Meeting ------------------------------ Date: Mon, 15 Jun 87 17:29:16 +0200 From: mcvax!cwi.nl!tomi@seismo.CSS.GOV (Tetsuo Tomiyama) Reply-to: mcvax!cwi.nl!tomi@seismo.CSS.GOV Subject: Conference - Second Eurographics Workshop on Intelligent CAD Systems Call For Papers SECOND EUROGRAPHICS WORKSHOP ON INTELLIGENT CAD SYSTEMS -Implementational Issues- APRIL 12-15, 1988, THE NETHERLANDS Organized by CENTRE FOR MATHEMATICS AND COMPUTER SCIENCE (CWI), AMSTERDAM Sponsored by EUROGRAPHICS AIM AND SCOPE This is the second workshop of a series of three Eurographics workshops on Intelligent CAD Systems which have the following main topics; - 1st, 1987: Theoretical and methodological aspects. - 2nd, 1988: Implementational issues. - 3rd, 1989: Practical experiences and evaluation. Since applying knowledge engineering to CAD systems seems very promising to solve the problems of conventional CAD systems, it has drawn attention from not only CAD researchers but also AI researchers. The first workshop which was held on April 21-24, 1987, in the Netherlands, aimed at discussing the results and problems in this highly interesting field. We have realized that ad hoc approaches will eventually result in increased complexity of CAD applications and that we need a robust theoretical basis for development. This second workshop in 1988 is planned to discuss implementational issues and to clarify problems associated with developing intelligent CAD systems based on those theoretical and methodological considerations. The scope of the workshop includes, but is not limited to; 1) Role of theories to implement intelligent CAD systems. 2) Implementations of theories for intelligent CAD systems. 3) Architecture of intelligent CAD systems. 4) Techniques and tools to implement intelligent CAD systems. 5) Acquisition and maintenance of design knowledge. 6) Innovative and large-scale implementations of intelligent CAD systems. 7) Problems and future tasks in implementations of intelligent CAD systems. We are especially interested in reports telling how theoretical work influenced implementations. SCHEDULE FOR THE SECOND WORKSHOP November 1, 1987: Deadline for extended abstracts and position papers. December 1987: Notification of acceptance for presentation. February 1988: Acceptance of participation. April 12-15, 1988: Workshop (Full papers are submitted just before the workshop). May 1988: Deadline for final manuscripts for publication. SERIES SCHEDULE Approximately 15 reviewed papers will be presented in this second workshop. Participants will be limited to about 50 based on invitation. Intended authors and participants are invited to submit extended abstracts or position papers. The results of this series of three workshops will be published by Springer-Verlag as Eurographics Seminar Books. The report on the first workshop held in April 1987, "Intelligent CAD Systems 1: Theoretical and Methodological Aspects," will be published in August 1987. This series of workshop is being organized under cooperation with IFIP Working Group 5.2 Workshops on Intelligent CAD Systems but with different scopes. ORGANIZATION Co-Chairmen P.J.W. ten Hagen (CWI, NL) T. Tomiyama (The University of Tokyo, J) Technical Secretary P.J. Veerkamp (CWI, NL) Workshop Secretary E. Both (CWI, NL) Program Committee A.M. Agogino (University of California, Berkeley, USA) V. Akman (CWI, NL) F. Arbab (University of Southern California, USA) P. Bernus (Hungarian Academy of Sciences, H) A. Bijl (University of Edinburgh, UK) J. Encarnacao (TH Darmstadt, D) S.J. Fenves (Carnegie Mellon University, USA) D. Gossard (MIT, USA) F. Kimura (The University of Tokyo, J) T. Kjellberg (Royal Institute of Technology, S) G.A. Kramer (Schlumberger Palo Alto Research, USA) M. Mac an Airchinnigh (University of Dublin, IRL) K. MacCallum (University of Strathclyde, UK) S. Murthy (IBM Thomas J. Watson Research, USA) F.J. Schramel (Philips, NL) D. Sriram (MIT, USA) W. Strasser (Universitaet Tuebingen, D) T. Takala (Technical University of Helsinki, SF) F. Tolman (TNO, NL) INFORMATION Please submit 5 copies of an extended abstract or a position paper up to 1,000 words (figures and references do not count) on A4 sheets before November 1, 1987, to: (Submission by electric mail is accepted) Ms. Elisabeth Both Centre for Mathematics and Computer Science Kruislaan 413, 1098 SJ Amsterdam, The Netherlands Telephone: (overseas) +31-20-592-4171 (from the Netherlands) 020-592-4171 Telex: 12571 mactr nl Electric Mail: pauljan@cwi.nl (Internet, Bitnet), ...!mcvax!pauljan (Usenet) The extended abstract or position paper should contain the following information: - Name, Address (Postal, Phone, Telex, E-mail), Keywords, References. - Statements on how you define "design" and "intelligent CAD systems." ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jun 18 16:56:15 1987 Date: Thu, 18 Jun 87 16:56:06 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #152 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 18 Jun 87 16:54 EDT Received: from relay.cs.net by RELAY.CS.NET id ac28701; 18 Jun 87 3:37 EDT Received: from [10.4.0.2] by RELAY.CS.NET id aa04521; 18 Jun 87 3:38 EDT Date: Thu 18 Jun 1987 00:07-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #152 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Thursday, 18 Jun 1987 Volume 5 : Issue 152 Today's Topics: Queries - Nano-Engineering & AI Research in Network Management & K-B Expert Systems for Manufacturing & Information Management in Software Engineering & Unix Lisps in C, AI Tools - ID3 vs C4 & Smalltalk-80 & ML, Comments - AI Models in Biology ---------------------------------------------------------------------- Date: 16 Jun 1987 09:10-EDT From: DAVSMITH@A.ISI.EDU Subject: Nano-Engineering The recent discussion of the $6M man reminded me of an oddity which someone out there in Net-land might be able to clarify. Early one morning on NPR (National Public Radio) I was surprised to hear a feature from someone at the MIT AI Lab entitled Nano-Engineering. I hasten to add that it was several months ago, but _not_ on April 1st, although the following synopsis may lead you to believe such. The general thesis was a genetic engineering exercise whereby a little genetic robot would be created to "assemble" genes. The really intersting part was the observation that since these things would naturally be very small, their first assignment would be to assemble clones of themselves. Recall that I said this was early in the morning, but I did check with another NPR fan in our office who also heard the same feature. Can anyone confirm (a) that this was perpetrated and (b) that it came from MIT? David Smith - DAVSMITH@A.ISI.EDU ------------------------------ Date: 16 Jun 87 21:33:30 GMT From: dvorak@im4u.utexas.edu (Daniel L. Dvorak) Subject: AI research in network management This is a brainstorming exercise, folks --- all ideas are welcome. I'm trying to select a PhD research topic in artificial intelligence that is applicable to network management (of data or voice networks) or, more liberally, the management of distributed computing environments. Network management, roughly, is concerned with the operation, administration and maintenance of communication networks, whether it be the campus network here at The University of Texas at Austin or the nationwide telephone network. The term encompasses issues such as congestion control, fault diagnosis, capacity planning, security, availability, etc. My questions for you are: -- What are the important unsolved (or poorly solved) problems here that might yield to AI? Please be specific. -- What AI research issues should be tested in this domain? -- Are there any papers that you would recommend to me? -- ----- Dan Dvorak UUCP: {harvard,ihnp4,seismo}!ut-sally!im4u!dvorak (512) 472-6671 ARPA: dvorak@im4u.utexas.edu ------------------------------ Date: 16 Jun 87 13:31:30 GMT From: pt!andrew.cmu.edu!dg1v+@cs.rochester.edu (David Greene) Subject: K-B Expert Sys for Manufacturing Could anyone tell me where I might obtain the following proceedings: Knowledge-Based Expert Systems for Manufacturing. Proceedings of the Winter Annual Meeting of the American Society of Mechanical Engineers (ASME), S. C-Y. Lu and R. Komanduri (eds.), Anaheim (CA), December 7-12, 1986. Please leave mail for: dg1v@andrew.cmu.edu David Greene GSIA Carnegie Mellon Univ. Pittsburgh, PA 15213 ------------------------------ Date: 16 Jun 87 17:41:17 GMT From: pollux.usc.edu!garg@OBERON.USC.EDU (Pankaj Garg) Subject: Information management in software engg. Hi, I am doing a survey on information management in the development, use, and maintenance of large scale software systems. I know about several efforts, but would like to be comprehensive, hence this posting. If you know of any efforts in databases, information science, or knowledge representation, in this direction, please let me know. I can post summaries to those interested. regards... ...pankaj US MAIL: Computer Science Department Sal 200 University of Southern California Los Angeles, CA 90089-0782 E-Mail: garg@pollux.usc.edu or garg@cse.usc.edu Phone: (213)743-7995, 735-2843 ------------------------------ Date: Wed, 17 Jun 87 13:21:22 BST From: A system manager Subject: Unix Lisps in C? I am seeking information about Unix lisps written entirely in (hopefully not VAX-specific) C. Pointers to such beasts would be gratefully received. Anybody who wants a copy of information thus obtained should let me know - I will be happy to forward it. Malcolm MacCallum (mm@maths.qmc.ac.uk) Relays: UKACRL (Bitnet), ucl-cs (arpa) ------------------------------ Date: 14 Jun 87 21:05:39 GMT From: mcvax!ukc!stc!praxis!gerry@seismo.css.gov (Gerry Wolff) Subject: Re: ID3 vs C4 In article <114@upba.UUCP> damon@upba.UUCP (Damon Scaggs) writes: >I understand that Ross Quinlan, author of the ID3 classification algorithm >has developed a better version with the designation C4. I am looking for >any papers or references about this new algorithm as well as any comments >about what it does better. I can't speak for C4 but I will claim, immodestly, that an inductive learning program I wrote (and reported) a few years ago is, in certain respects, more sophisticated than ID3. In particular, it integrates the learning of segmental structure with the learning of disjunctive (class) structure. The program (called SNPR) also has the ability to generalize structures and to correct overgeneralizations *without correction by a 'teacher' or the provision of 'negative' samples*. The reference is: Wolff J G (1982). Language acquisition, data compression and generalization. Language & Communication 2, 57-89. *-----------------------------------------------------------------------* | Dr Gerry Wolff | Phone: (44) 225 335855 | | Praxis Systems plc | UUCP: gerry@praxis.co.uk | | 20 Manvers Street | Telex: 445848 PRAXIS G | | Bath | Facsimile Groups 2 & 3 | | BA1 1PX | (44) 225 65205 | | UK | | *-----------------------------------------------------------------------* ------------------------------ Date: Mon, 15 Jun 87 16:04:21 PDT From: "William J. Fulco" Subject: Smalltakl-80 for Sun 3 (and others) I saw a really nice system, (I mean REALLY nice - with good color support) from Xerox PARC marketing spinoff at the 1986 AAAI show. It was running on a Sun 3/260 and it really sizzles..... I believe they are going to show this and some other versions (a la Mac) at the U of WA/Seattle AAAI '87 show July. I'll be there, Mac II in hand, drooling !!!!! I have been waiting (2 months) paitently for information from: ParcPlace Systems 3330 Coyote Hill Road Palo Alto, CA 94304 (415) 859-1000 (bill) P.S. I would be interested in any implementations you find out about. ------------------------------ Date: 16 Jun 87 01:14:50 GMT From: mcvax!diku!carllp@seismo.css.gov (Carl-Lykke Pedersen) Subject: Re: ML programming, anybody? Yes, we are some people at diku (datalogisk institut ved K|benhavns Universitet -> Coputer Science Department at the University of Copenhagen) who are trying to work with SML. We are supposed to make a user-manual for the implementation - but .... Right now I'm working with a self-interpreter to SML, and it seems to be ok. We are using a version from Edingburgh. It's rather old, but we have some problems in getting a newer version. Regards Carl-Lykke ------------------------------ Date: 11 Jun 87 16:10:31 GMT From: ptsfa!hoptoad!academ!uhnix1!uhnix2!bchso@ames.arpa (Dan Davison) Subject: Re: Taking AI models and applying them to biology... In article <1331@sigi.Colorado.EDU> pell@boulder.Colorado.EDU (Anthony Pelletier) writes: >P.S. I think alot about information flow problems and would enjoy >discussions on that...if anyone wants to chat. Do you know about the "Matrix of Biological Knowledge Workshop" in Santa Fe, NM July 13-August 14 this year? One of the subjects is "information flow from DNA to cells" lead by Dickerson of UCLA, Hershman, also UCLA, and Smith from MBCRR at Harvard. For information, contact Ms. Ginger Richardson at The Santa Fe Institute, P.O. Box 9020, Santa Fe, N.M. 87504-9020; phone 505-984-8800. dr. dan davison/ Dept of Biochemical and Biophysical Sciences/ U. of Houston bitnet: bchs6\@uhupvm1.bitnet | 4800 Calhoun/ Houston, Tx 77004 arpanet: davison\@sumex-aim.stanford.edu|uucp:...rice!academ!uhnix1!uhnix2!bchso ------------------------------ Date: 15 Jun 87 13:06:19 GMT From: edwards@unix.macc.wisc.edu (mark edwards) Subject: Re: Taking AI models and applying them to biology... In article <7416@boring.cwi.nl> lambert@cwi.nl (Lambert Meertens) writes: :| > This description of the human memory system, though cloaked in vaguer terms, :| > corresponds more or less one-to-one with the traditional computer :| > architecture we all know and love. To wit: :| :| [description deleted] :| :| > At least this far, this theory appears to owe a lot to computer science. :| > Granted, there is lots of empirical evidence in favour, but we all know :| > how a little evidence can go far too far towards developing an analogy. :The main similarity appears to be that several levels of memory can be :discerned, but the suggested analogy in function is a bit far-fetched. : :It is perhaps worth pointing out that much of the current models in :cognitive psychology can already be found in the pioneering work of Otto :Selz (Muenchen, 1881 - Auschwitz, 1943), antedating the computer era. What? You cite facts from the pre-computer age? Shame shame. Don't you know that life began with the creation of the computer, as well as all the other sciences! With out the computer all other life would cease to exist! Its a sad fact that the above holds true for many computer scientist especially those in AI. Many still believe the sacred words AI were really coined in the late fifties, and that LISP and Liebnitz have no connection. When infact my prof. has given references from a latin book dating back to the 13th century, with the latin words for AI. The state of the art of computer science is in bad shape when the computer wheel must be re-invented every year because th CS people refuse to read any book that seemingly has nothing to do with CS or computers. Thanks for the reference. I may be the only one who benefits from it, because those CS'ers practicing the art would certainly declare it blasphemous and maybe just short of heresy. I may be stoned for this..... mark -- edwards@unix.macc.wisc.edu {allegra, ihnp4, seismo}!uwvax!uwmacc!edwards UW-Madison, 1210 West Dayton St., Madison WI 53706 ------------------------------ Date: 15 Jun 87 03:47:26 GMT From: ihnp4!alberta!mnetor!utzoo!utgpu!utcsri!utegc!utai!tjhorton@ucbv ax.Berkeley.EDU Subject: Re: Taking AI models and applying them to biology... >lambert@cwi.nl (Lambert Meertens) writes: >It is perhaps worth pointing out that much of the current models in >cognitive psychology can already be found in the pioneering work of Otto >Selz (Muenchen, 1881 - Auschwitz, 1943), antedating the computer era. 1943 was at least 7 years after Turing published his paper (fifty years ago, last November) and 5 years after Shannon published his thesis about information theory. Although I don't know Selz, his life definitely spanned into the dawn of the "computer era". It's interesting - do these models of his pre-date these "computeresque" notions? Timothy J Horton ------------------------------ Date: 16 Jun 87 18:04:06 GMT From: pyramid!prls!philabs!aecom!diaz@decwrl.dec.com (Dizzy Dan) Subject: Re: Taking AI models and applying them to biology... In article <395@uhnix2.UUCP>, bchso@uhnix2.UUCP (Dan Davison) writes: > Do you know about the "Matrix of Biological Knowledge Workshop" in Santa Fe, > NM > July 13-August 14 this year? One of the subjects is "information flow from > DNA to cells" lead by Dickerson of UCLA, Hershman, also UCLA, and Smith from > MBCRR at Harvard. > > For information, contact Ms. Ginger Richardson at The Santa Fe Institute, > P.O. Box 9020, Santa Fe, N.M. 87504-9020; phone 505-984-8800. > Sorry gang, but applications for the Matrix Workshop were due in April. If you are interested, the Santa Fe Institute may be able to put you in touch with some of the faculty running the workshops. -- 5'gtacggagc dn/dx = Dan Diaz (philabs!aecom!diaz) Department of Molecular Biology & Snake Oil Dynamics Albert Slimestein College of Medicine ctataacagcta 3' ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Sat Jun 20 03:24:19 1987 Date: Sat, 20 Jun 87 03:24:13 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #153 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Sat, 20 Jun 87 03:20 EDT Received: from relay.cs.net by RELAY.CS.NET id aa11864; 20 Jun 87 1:33 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa26357; 20 Jun 87 1:27 EDT Date: Fri 19 Jun 1987 22:02-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #153 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Saturday, 20 Jun 1987 Volume 5 : Issue 153 Today's Topics: Queries - Computer Composition of Music & AI in Criminal Investigation & Expert Systems in Process Control, AI Tools - Unix LISPs in C, Speculation - Nanotechnology, Psychology - Why Did The $6,000,000 Man Run So Slowly? ---------------------------------------------------------------------- Date: 17 Jun 87 19:50:07 GMT From: pwa-b!mmintl!johnt@gr.utah.edu (John Tangney) Subject: Computer composition of music Composing music -- procedurally: Until 1981 I had been doing some research on computer composition of music. I have now taken up where I left off. Even back then I was not up to date with the latest advances. A lot must have happened in this field since then. I want to contact people who know something about composing by computer. Some of the researchers I read about (like Max Mathews, Lejaren Hiller, Iannis Xenakis, Stephen Smoliar to name a few off the top of my head) must still be out there. Also, people like myself, who do this in their spare time, must have ideas, suggestions, sources of information. What about journals, books, papers on the subject? If anyone in net.land knows anything about computer composition of music, or knows anyone who does, then I beg you to let me know. E-mail is most sensible. Who knows? Maybe we could end up with our own news.group! A phone call or Snailmail reply would be most welcome too. John Tangney ...inhp4!philabs!pwa-b!mmintl!johnt 52 Oakland Ave, East Hartford, CT 06108. Phone: (203) 522-2116 ------------------------------ Date: 19 Jun 87 08:21 PDT From: gaska.pasa@Xerox.COM Subject: Use of AI in Criminal Investigation Does any one have any pointers to papers, books, persons, etc. that deal with the use of AI in criminal investigation and forensic science? Any and all leads will be greatly appreciated. Len Gaska GASKA@PASA.XEROX.COM ------------------------------ Date: 18 Jun 1987 11:58:37 EST From: Herve.Lambert@PS3.CS.CMU.EDU Subject: Expert Systems in Process Control I have to find some literature about actually operational expert systems in process control. All I know is the example of PICON used at a Texaco's refinery. Any informations, pointers very much aprreciated... If I get interesting enough info, and if some people express the desire to have the result of my query posted, I will do so... Thanks in advance - Herve Net-address: herve@ps3.cs.cmu.edu [How about this blurb from Business Week, "The 'Renaissance Man' of Expert Systems?", Emily T. Smith, May 11, 1978, p. 141: The trouble with using so-called expert system computer programs in the factory is that very few manufacturing operations involve only one realm of expertise. It's tough enough getting two experts in the same field to agree, let alone a gaggle of experts from different disciplines. So Major Stephen R. LeClair, head of research in artificial intelligence for manufacturing at the Materials Laboratory at Wright-Patterson Air Force Base, decided it was time to devise a new type of expert system -- one that could embrace multiple "domains" of expertise, automatically resolve any conflicts, and "learn" from the process. In its first real-world test, LeClair's multiexpert knowledge system (MKS) recently turned in a stunning performance. It discovered, on its own, that the accepted guidelines for curing complex plastics composites are all wet. The aerospace industry has been taking 12 hours to bake a 256-layer, graphite-reinforced lamination used for airframe parts. By synthesizing the knowledge from various fields, MKS came up with a complicated scheme for curing the composite in less than three hours. No one believed it could work, but it does. LeClair asserts that MKS may similarly confound convential wisdom in other process-control applications. The same page has another short report about a system that measures rough gemstones (other than diamonds), plans the optimal cuts, and cuts the stones. It reduces wasted stone by 10%, cost by 70%, and makes marginal stones useable. -- KIL] ------------------------------ Date: Fri, 19 Jun 87 12:11:16 EDT From: dml@nadc.arpa (D. Loewenstern) Subject: Unix LISPs in C In response to your request for Unix LISPs written entirely in C, I believe I can recommend Kyoto Common Lisp. It has no real editor (it shells out to vi!!) but it implements nearly the entirety of Common LISP. The compiler translates to C, then invokes the C compiler. I know of VAX, ECLIPSE, and Sun versions. Write to: IBUKI 399 Main Street Los Altos, CA 94022 David Loewenstern Naval Air Development Center code 7013 Warminster, PA 18974-5000 ------------------------------ Date: Thu 18 Jun 87 17:21:23-CDT From: Jonathan Slocum Subject: nanomachinery The book "Engines of Creation," by one K. Eric Drexler, describes this technology and discusses the societal ramifications of its introduction. Whether one believes in the possibility of such things or not (Drexler is a persuasive advocate), it makes for good reading in my opinion. He was (is?), I believe, associated with MIT in some way (don't have the book with me, so can't refer to the jacket) -- perhaps a student?? -Jonathan Slocum ------------------------------ Date: 18 Jun 87 07:35:44 edt From: Walter Hamscher Subject: Nano-Engineering Date: 16 Jun 1987 09:10-EDT From: DAVSMITH@A.ISI.EDU The recent discussion of the $6M man reminded me of an oddity which someone out there in Net-land might be able to clarify. Early one morning on NPR (National Public Radio) I was surprised to hear a feature from someone at the MIT AI Lab entitled Nano-Engineering. * * * Can anyone confirm (a) that this was perpetrated and (b) that it came from MIT? Its proponents call it Nanotechnology. The most well known spokesman seems to be Eric Drexler, who has written a book about it called "Engines of Creation." I think it's from MIT Press. Below I have included an announcement of a two day symposium that was held during IAP (Independent Activities Period, known as "January" to the world outside MIT). As you can see from the header of the message, there is a mailing list called nanotechnology@oz.ai.mit.edu, or, from outside MIT a better bet would be to try nanotechnology@ai.ai.mit.edu. ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; Date: Fri, 16 Jan 87 02:35 EST From: Christopher Fry Subject: Nanotechnology Symposium To: nanotechnology@OZ.AI.MIT.EDU, MACROMOLECULES-MIT@OZ.AI.MIT.EDU, ROBOTICS-SEMINARS@OZ.AI.MIT.EDU, *BBOARD@OZ.AI.MIT.EDU Exploring Nanotechnology An IAP 87 Symposium All technology rests ultimately on our ability to arrange atoms. Foreseeable technological advances will enable us to build devices to atomic specifications. This "nanotechnology" will have profound consequences, forcing a reevaluation of our expectations regarding the next several decades. In the symposium, we will explore paths to the development of nanotechnology, consequences of the technology in various disciplines, and we will critically examine the premises of these assertions via panel discussions which will include experts in several fields. Tuesday, 20 January 1987, 10-250 10:00 - 11:00 am Overview: Eric Drexler (BS '77, MS '79) will describe various paths to the development of replicating assembler systems, capable of manufacturing complex components to atomic specification. Some potential applications, such as mechanical nanocomputers, and their consequences will be discussed. We strongly recommend you attend this talk in order to follow the subsequent discussions in context. 11:05 - 11:45 am Materials Science and Protein Engineering: Kevin Ulmer will discuss the protein engineering techniques which could be used to create new alloys and composites. New materials made to atomic specifications promise order of magnitude improvements in performance. One consequence is space transportation costs equivalent to current airline costs. Noon - 1:00 pm Lunch Break 1:00 - 1:40 pm Panel Discussion I. A panel of experts will discuss the technical feasibility of various aspects of nanotechnology, including consideration of the time frame. A panel moderator will take questions from the audience. 1:45 - 2:25 pm Economics: David Friedman will discuss the consequences of nanotechnology, such as extreme decentralization of the economy. On-site, personal manufacturing stations could virtually eliminate mass production. What will happen to our economy during the transition to this technology? 2:30 - 3:10 pm Society, Technology and Policy: Arthur Kantrowitz will share his thoughts on how society may be affected, and what kind of future may be in store for the human race. How can our government adapt to this new technology and what legislation, if any, should be enacted to control its development? 3:10 - 3:25 Break. 3:25 - 4:05 pm Thought and Intelligence: Marvin Minsky will speak on intelligent systems which could employ Avogadro's number of parallel nanocomputers. Achieving artificial intelligence by mimicking human brain architecture is a rapid route to true AI with nanotechnology. 4:10 - 4:40 pm Concluding Points: Eric Drexler will wrap up by describing life extension possibilities using cell repair machines. 4:10 - 5:00 pm Panel Discussion II. A panel of experts will discuss the societal implications of nanotechnology, including steps we might take to avoid some of the dangerous consequences of nanotechnology. A panel moderator will take questions from the audience. Thursday, 22 January 1987 7:30 - 10:00 pm Advanced Topics: NE43-773 As an extension to the symposium we will hold a special session during the regular meeting time of the MIT Nanotechnology Study Group. We will discuss, in depth, critical issues regarding the development of nanotechnology such as control of assemblers, guidance of technology development, and prevention of abuse. Eric Drexler will be with us. Recommended only for those who attend the symposium on Tuesday, or who have attended NSG introductory lectures in the past. Sponsored by the MIT Nanotechnology Study Group, the Dept. of Applied Biological Sciences, the Artificial Intelligence Laboratory, the Office of the Associate Provost, the Graduate Student Council, the Dept. of Materials Science and Engineering, and the Dept. of Political Science. Special thanks to the AI Lab for its generous support of this activity. Contact cfry@@MIT-OZ ------------------------------ Date: 18 Jun 87 08:53:01 CDT (Thu) From: ernst%home%ti-csl.csnet@RELAY.CS.NET Subject: Nanotechnology The "nano-engineering" that David Smith heard about on an early-morning NPR show is, indeed, no joke. Its chief proponent is K. Eric Drexler, who describes the theory in his 1985 book _Engines of Creation_. He is associated with MIT, and he has a stong following there. In particular, Marvin Minsky wrote the forward to the above-mentioned book and spoke, along with Eric Drexler and others, at a recent day-long symposium on nano-technology at MIT. The idea behind nanotechnology is the creation of tiny machines which would be built up molecule by molecule by "molecular assemblers", which would function much like DNA or RNA in fishing for the right component to add to a structure. Because of their small size, their manipulators would move a million times a second, resulting in extraordinarily quick construction. Mechanical nanocomputers (that is, they would contain tiny gears made of a handful of atoms and such, on the order of Hillis's mechanical tic-tac-toe player) orders of magnitude more powerful than current machines but small enough to fit in dust-speck sized nanomachines would carry instructions and direct work. Drexler envisions the construction of assemblers within a few decades as a result of advances in bioengineering and other sciences. It is a technology he believes will powerfully leverage off itself: after the first assembler is built, uncountable trillions more will follow almost immediately, and scientific breakthroughs in many fields (all of which will be able to use nanotechnology or its products as a tool) will be made in days rather than years. Drexler's book is more about what changes will be made in society with the advent of nanomachines than their technical aspects; after all, no one is close to the advances he envisions. He discusses jet engines built in hours, self-repairing machines, AI workstations of unprecedented power, and world hypertext systems as well as more sinister possibilities like the capability to build tiny airborne surveillance devices or supergerms that could destroy life on earth in hours. Although much of the material is hard to believe, I recommed the book for an interesting mix of philosophy and forward-sighted scientific thought (or science fiction, call it what you like). -Michael Ernst MIT AI Lab Texas Instruments AI Lab mernst@oz.ai.mit.edu ernst%home%ti-csl@csnet-relay.arpa ...!eddie!mernst The opinions expressed above are not only not those of my employer, they may well not be my own. ------------------------------ Date: Mon, 15 Jun 87 14:48 EDT From: Nichael Cramer Reply-to: Nichael Cramer Subject: Re: Why Did The $6,000,000 Man Run So Slowly? >> >> Date: Fri, 12 Jun 87 00:51:41 EDT >> From: tim@linc.cis.upenn.edu (Tim Finin) >> Subject: why did the $6,000,000 man run so slowly? >> I had always assumed that he ran slowly for the same reason that the people on "Kung Fu" always fought so slowly; namely that it's technically much easier to depict graphic, physical motion (and violence) in this way. You have the first actor throwing punches that actually connect with the second actor's jaw, except that he's moving more slowly in real time, and so not crippling the second actor with every blow. Once you slow this down a lot, the viewer loses all sense of how much the time is really altered; i.e. the slow motion camera technique masks the slowed down "acting". In the present case, slow motion has the effect of distorting your time sense and allowing the film makers to use other (cheaper?) methods to suggest high-speed to the viewer, e.g. swooshing sounds or tense music. (With regard to these non-visual cues used to suggest high-speed: As others have pointed out, watch the opening of "Star Trek" with the sound turned off. The Enterprise, which would normally sweeping across the screen at Warp N, will, with the usual swooshing sound missing, simply creep across the screen.) NICHAEL ------------------------------ Date: 15 Jun 87 19:36:28 GMT From: tektronix!teklds!zeus!bobr@ucbvax.Berkeley.EDU (Robert Reed) Subject: Re: Why did the six-million dollar man run so slowly? Because it was cheaper to take slow motion footage to show SOMETHING was happening than to make a believable high speed effect. Of course, they could taken blue screen shots of Steve Austin running normally and composited in a high speed background, but many of the shots involved his feet. Making a believable shot under those conditions would have been a lot more expensive. It is interesting to note that the recent reunion of the "bionic family" represented the new generation of bionic technology by having his son blur (it actually looked like defocused multiple exposures) during the slow motion "high speed" running shots. -- Robert Reed, Tektronix CAE Systems Division, bobr@zeus.TEK ------------------------------ Date: Mon, 15 Jun 87 17:15:04 edt From: amsler@flash.bellcore.com (Robert Amsler) Subject: Re: AIList Digest V5 #145 Incidentally.... Re: Dr. Who's TARDIS. I've decided most of the discussions were wrong. Few people considered the function of the `relative dimensions stabilizer circuits' which are intended to compensate for dimensional anomalies. It would be QUITE possible to have the inside view of the TARDIS look either miniturized or like a small window into a larger room. One should recall that anomalies in the circuit can cause the TARDIS inhabitants to actually BE smaller when they emerge. Anyway... wrong discussion. `pop' Re: bionics. It has been my belief for some time that the mind operates using movie techniques when examining moving image memories. That is, we employ cuts, zooms, view angles, props, etc. in such memory recording and dreams. It would seem reasonable that we have borrowed this acceptable form of imaging and used it in films--why, for instance, should a cut between two views be acceptable cinamatography. Some cinematographic techniques violate our `dream' view methods. For instance, when one holds the camera at a bad angle the impact is typically to introduce the concept of the camera into the film, i.e. one way to show something is being seen through a camera lens in a film, is to have the camera do bad cinematographic techniques--ones which make the artificiality of the instrument apparent (another problem is whenever things get on the lens, such as rain or ocean spray or dust, etc.) Now, the speed to slow motion effect is interesting in that I don't believe it does have a natural human moving image memory counterpart. We never see things in slow motion ourselves, except as they have been slowed down by the use of film etc. That indeed explains to me why this is being discussed in AILIST. I.e. it is an artificial learned moving-image association. The interesting thing is that is SEEMS to be possible to introduce this into the visual recording system for memories in the brain without causing the ``Oh, this is being shot through a camera'' phenomena. I suspect what is happening is that this is analogous to the focusing of attention on the events which happened in a real moving image memory. That is, if one attempts to reconstruct an event that happened very quickly in real time after the fact, one will artificially create something like slow motion. ---- Note: I am NOT saying that we really have moving images in the brain. It is unclear we have images at ALL; however, the mapping between what we do have and what we accept in cinematographic portrayals is an interesting one. ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Sun Jun 21 03:31:16 1987 Date: Sun, 21 Jun 87 03:31:06 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #154 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Sun, 21 Jun 87 03:26 EDT Received: from relay.cs.net by RELAY.CS.NET id aa02085; 20 Jun 87 19:39 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa02779; 20 Jun 87 19:25 EDT Date: Sat 20 Jun 1987 16:08-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #154 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Sunday, 21 Jun 1987 Volume 5 : Issue 154 Today's Topics: Theory - Symbol Grounding and Physical Invertibility ---------------------------------------------------------------------- Date: 16 Jun 87 1559 PDT From: John McCarthy Subject: Symbol Grounding Problem and Disputes [In reply to message sent Mon 15 Jun 1987 23:23-PDT.] This dispute strikes me as unnecessarily longwinded. I imagine that the alleged point at issue and a few of the positions taken could be summarized for the benefit of those of us whose subjective probability that there is a real point at issue is too low to motivate studying the entire discussion but high enough to motivate reading a summary. ------------------------------ Date: 16 Jun 87 17:41:50 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem (Reply to Ken Laws on ailist) In article <849@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > .... Invertibility could fail to capture the standard A/D distinction, > but may be important in the special case of mind-modeling. Or it could > turn out not to be useful at all.... So what do you think is essential: (A) literally analog transformation, (B) invertibility, or (C) preservation of significant relational functions? > ..... what I've said about the grounding problem and the role > of nonsymbolic representations (analog and categorical) would stand > independently of my particular criterion for analog; substituting a more > standard one leaves just about all of the argument intact..... Where does that argument stand now? Can we restate it in terms whose definitions we all agree on? > ..... to get the requisite causality I'm looking > for, the information must be interpretation-independent. Physical > invertibility seems to give you that...... I think invertibility is too strong. It is sufficient, but not necessary, for human-style information-processing. Real people forget awesome amounts of detail, we misunderstand each other (our symbol groundings are not fully invertible), and we thereby achieve levels of communication that often, but not always, satisify us. Do you still say we only need transformations that are analog (invertible) with respect to those features for which they are analog (invertible)? That amounts to limited invertibility, and the next essential step would be to identify the features that need invertibility, as distinct from those that can be thrown away. > Ken Laws on ailist@Stripe.SRI.Com writes: > > ... I am sure that methods for decoding both discrete and > > continuous information in continuous signals are well studied. > > I would be interested to hear from those who are familiar with such work. > It may be that some of it is relevant to cognitive and neural modeling > and even the symbol grounding problems under discussion here. I'm not up to date on these methods. But if you want to get responses from experts, it might be well to be more specific. For monaural sound, decoding can be done with Fourier methods that are in principle continuous. For monocular vision, Fourier methods are used for image enhancement to aid in human decoding, but I think machine decoding depends on making the spatial dimensions discontinous and comparing the content of adjacent cells. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: Wed 17 Jun 87 23:33:01-PDT From: Ken Laws Subject: Visual Decoding From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) For monaural sound, decoding can be done with Fourier methods that are in principle continuous. For monocular vision, Fourier methods are used for image enhancement to aid in human decoding, but I think machine decoding depends on making the spatial dimensions discontinous and comparing the content of adjacent cells. Marty is right; one must be specific about the types of signals that are carrying the information. Information theorists tend to work with particular types of modulation (e.g., radar returns), but are interested in the general principles of information transmission. Some of the spread spectrum work is aimed at concealing evidence of modulation while still being able to recover the encoded information. Fourier techniques are particularly appropriate for speech processing because sinusoidal waveforms (the basis of Fourier analysis) are the eigenforms of acoustic channels. In other words, the sinusoidal components of speech are transmitted relatively unharmed, although the phase relationships between the components can be scrambled. Any process that decodes acoustic signals must be prepared to deal with a little phase spreading. Other 1-D signals (e.g., spectrographic signatures of chemicals) may be composed of Gaussian pulses or other basis forms. Yet others may be generated by differential equations rather than composition or modulation of basis functions. Decoding generally requires models of the generating process and of the channel or sensing transformations, particularly if the latter are invertible. Images are typically captured in discrete arrays, although we know that biological retinas are neither limited to one kind of detector/resolution nor so spatially regular. Discrete arrays are convenient, and the Nyquist theorem (combined with the limited spatial resolution of typical imaging systems) gives us assurance that we lose nothing below a specific minimum frequency -- we can, if we wish, reconstruct the true image intensity at any point in the image plane, regardless of its relationship to the pixel centers. (In practice this interpolation is exceedingly difficult and is almost never done -- but enough pixels are sampled to make interpolation unnecessary for the types of discrimination we need to perform.) The discrete pixel grid is often convenient but is not fundamental to the enterprise of image analysis. A difficulty in image analysis is that we rarely know the shapes of the basis functions that carry the information; that, after all, is what we are trying to determine by parsing a scene into objects. We do have models of the optical channels, but they are generally noninvertible. Our models of the generating processes (e.g., real-world scenes) are exceedingly weak. We have some approaches to decoding these signals, but nothing approaching the power of the human visual system except in very special tasks (such as analysis of bubble chamber photographs). -- Ken ------------------------------ Date: 17 Jun 87 08:02:00 EST From: cugini@icst-ecf.arpa Reply-to: Subject: symbol grounding and physical invertibility I hate to nag but... In all the high-falutin' philosophical give-and-take (of which, I admit, I am actually quite fond) there's been no response to a much more *specific* objection/question I raised earlier: What if there were a few-to-one transformation between the skin-level sensors (remember Harnad proposes "skin-and-in" invertibility as being necessary for grounding) and the (somewhat more internal) iconic representation. My example was to suppose that #1: a combination of both red and green retinal receptors and #2 a yellow receptor BOTH generated the same iconic yellow. Clearly this iconic representation is non-invertible back out to the sensory surfaces, but intuitively it seems like it would be grounded nonetheless - how about it? John Cugini ------------------------------ Date: 17 Jun 87 18:32:20 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem (Reply to Ken Laws on ailist) In article <849@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes: > As long as the requisite >information-preserving mapping or "relational function" is in the head >of the human interpreter, you do not have an invertible (hence analog) >transformation. But as soon as the inverse function is wired in >physically, producing a dedicated invertible transformation, you do >have invertibility, ... This seems to relate to a distinction between "physical invertibility" and plain old invertibility, another of your points which I haven't understood. I don't see any difference between "physical" and "merely theoretical" invertibility. If a particular physical transformation of a signal is invertible in theory, then I'd imagine we could always build a device to perform the actual inversion if we wanted to. Such a device would of course be a physical device; hence the invertibility would seem to count as "physical," at least in the sense of "physically possible". Surely you don't mean that a transformation-inversion capability must actually be present in the device for it to count as "analog" in your sense. (Else brains, for example, wouldn't count). So what difference are you trying to capture with this distinction? Anders Weinstein BBN Labs ------------------------------ Date: 17 Jun 87 20:12:22 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem marty1@houdi.UUCP (M.BRILLIANT) asks: > what do you think is essential: (A) literally analog transformation, > (B) invertibility, or (C) preservation of significant relational > functions? Essential for what? For (i) generating the pairwise same/different judgments, simlarity judgments and matching that I've called, collectively, "discrimination", and for which I've hypothesized that there are iconic ("analog") representations? For that I think invertibility is essential. (I think that in most real cases what is actually physically invertible in my sense will also turn out to be "literally analog" in a more standard sense. Dedicated digital equivalents that would also have yielded invertibility will be like a Rube-Goldberg alternative; they will have a much bigger processing cost. But for my puroposes, the dedicated digital equivalent would in principle serve just as well. Don't forget the *dedicated* constraint though.) For (ii) generating the reliable sorting and labeling of objects on the basis of their sensory projections, which I've called collectively, "identification" or "categorization"? For that I think only distinctive features need to be extracted from the sensory projection. The rest need not be invertible. Iconic representations are one-to-one with the sensory projection; categorical representations are many-to-few. But if you're not talking about sensory discrimination or about stimulus categorization but about, say, (iii) conscious problem-solving, deduction, or linguistic description, then relation-preserving symbolic representations would be optimal -- only the ones I advocate would not be autonomous (modular). The atomic terms of which they were composed would be the labels of categories in the above sense, and hence they would be grounded in and constrained by the nonsymbolic representations. They would preserve relations not just in virtue of their syntactic form, as mediated by an interpretation; their meanings would be "fixed" by their causal connections with the nonsymbolic representations that ground their atoms. But if your question concerns what I think is nesessary to pass the Total Turing Test (TTT), I think you need all of (i) - (iii), grounded bottom-up in the way I've described. > Where does [the symbol grounding] argument stand now? Can we > restate it in terms whose definitions we all agree on? The symbols of an autonomous symbol-manipulating module are ungrounded. Their "meanings" depend on the mediation of human interpretation. If an attempt is made to "ground" them merely by linking the symbolic module with input/output modules in a dedicated system, all you will ever get is toy models: Small, nonrepresentative, nongeneralizable pieces of intelligent performance (a valid objective for AI, by the way, but not for cognitive modeling). This is only a conjecture, however, based on current toy performance models and the the kind of thing it takes to make them work. If a top-down symbolic module linked to peripherals could successfully pass the TTT that way, however, nothing would be left of the symbol grounding problem. My own alternative has to do with the way symbolic models work (and don't work). The hypothesis is that a hybrid symbolic/nonsymbolic model along the lines sketched above will be needed in order to pass the TTT. It will require a bottom-up, nonmodular grounding of its symbolic representations in nonsymbolic representations: iconic ( = invertible with the sensory projection) and categorical ( = invertible only with the invariant features of category members that are preserved in the sensory projection and are sufficient to guide reliable categorization). > I think invertibility is too strong. It is sufficient, but not > necessary, for human-style information-processing. Real people > forget... misunderstand... I think this is not the relevant form of evidence bearing on this question. Sure we forget, etc., but the question concerns what it takes to get it right when we actually do get it right. How do we discriminate, categorize, identify and describe things as well as we do (TTT-level) based on the sensory data we get? And I have to remind you again: categorization involves at least as much selective *non*invertibility as it does invertibility. Invertibility is needed where it's needed; it's not needed everywhere, indeed it may even be a handicap (see Luria's "Mind of a Mnemonist," which is about a person who seems to have had such vivid, accurate and persisting eidetic imagery that he couldn't selectively ignore or forget sensory details, and hence had great difficulty categorizing, abstracting and generalizing; Borges describes a similar case in "Funes the Memorious," and I discuss the problem in "Metaphor and Mental Duality," a chapter in Simon & Sholes' (eds.) "Language, Mind and Brain," Academic Press 1978). > Do you still say [1] we only need transformations that are analog > (invertible) with respect to those features for which they are analog > (invertible)? That amounts to limited invertibility, and the next > essential step would be [2] to identify the features that need > invertibility, as distinct from those that can be thrown away. Yes, I still say [1]. And yes, the category induction problem is [2]. Perhaps with the three-level division-of-labor I've described a connectionist algorithm or some other inductive mechanism would be able to find the invariant features that will subserve a sensory categorization from a given sample of confusable alternatives. That's the categorical representation. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: Thu, 18 Jun 87 10:08:27 pdt From: Ray Allis Subject: The Symbol Grounding Answer I have enjoyed the ailist's tone of rarified intellectual inquiry, but lately I have begun to think the form of the question "What is the solution to the Symbol Grounding Problem" has unduly influenced the content of the answer, as in "How many angels can dance on the head of a pin?" You are solemnly discussing angels and pinheads. There is no "Symbol Grounding Problem"; the things are *not* grounded. The only relationship a symbol has with anything is that the physical effects (electrical and chemical) of its perception in the brain of a perceiver co-exist with the physical effects of other perceptions, and are consequently associated in that individual's brain, and therefore mind. It happens when we direct our baby's attention at a bovine and clearly enunciate "COW". There is no more "physical invertibility" in that case than there is between you and your name, and there is no other physical relationship. And, as we computer hackers are wont to say, "That's a feature, not a bug". It means we can and do "think" about things and relationships which may not "exist". (BTW, it's even better! You are right now using second-level symbols. The visual patterns you are perceiving on paper or on a display screen are symbols for sounds, which in turn are symbols for experiences.) Last year's discussion of the definitions of "analog" and "digital" are relevant to the present topic. In the paragraph above, the electrical and chemical effects in the observer's brain are an *analogy* (we hypothesize) of external "reality". These events are *determined* (we believe) by that reality, i.e., for each external situation there is one and only one electro-chemical state of the observer's brain. Now, the brain effects appear pretty abstracted, or attenuated, so "complete invertibility" is unlikely, but if we can devise a fancy enough brain, may be approachable. No such deterministic relationship holds between external "reality" and symbols. As I noted above, symbols are related to their referents by totally arbitrary association. Thus, there is nothing subtle about the distinction between "analog" and "digital"; they are two profoundly different things. The "digital" side of an A/D relationship is *symbolic*. The relationship (we humans create) between a symbol and a quantity is wholly arbitrary. The value here is that we can use *deductive* relationships in our manipulation of quantities, rather than, say, pouring water back and forth among a set of containers to balance our bank account. I am one of those convinced by such considerations that purely symbolic means, which includes most everything we do on digital computers, are *insufficient in principle* to duplicate human behavior. And I have some ideas about the additional things we need to investigate. (By the way, whose behavior are we to duplicate? Ghengis Khan? William Shakespeare? Joe Sixpack? All of the above in one device? The Total Turing Test is just academic obfuscation of "If it walks like a duck, and quacks like A duck ..."). ------------------------------ Date: 18 Jun 87 18:26:23 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In article <861@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes: > The atomic terms of which they were >composed would be the labels of categories in the above sense, and hence they >would be grounded in and constrained by the nonsymbolic representations. >They would preserve relations not just in virtue of their syntactic >form, as mediated by an interpretation; their meanings would be "fixed" >by their causal connections with the nonsymbolic representations that >ground their atoms. I don't know how significant this is for your theory, but I think it's worth emphasizing that the *semantic* meaning of a symbol is still left largely unconstrained even after you take account of it's "grounding" in perceptual categorization. This is because what matters for intentional content is not the objective property in the world that's being detected, but rather how the subject *conceives* of that external property, a far more slippery notion. This point is emphasized in a different context in the Churchland's BBS reply to Drestke's "Knowledge and the Flow of Information." To paraphrase one of their examples: primitive people may be able to reliably categorize certain large-scale atmospheric electrical discharges; nevertheless, the semantic content of their corresponding states might be "Angry gods nearby" or some such. Indeed, by varying their factual beliefs we could invent cases where the semantic content of these states is just about anything you please. Semantic content is a holistic matter. Another well-known obstacle to moving from an objective to an intentional description is that the latter contains an essentially normative component, in that we must make some distinction between correct and erroneous classification. For example, we'd probably like to say that a frog has a fly-detector which is sometimes wrong, rather than a "moving-spot-against-a- fixed-background" detector which is infallible. Again, this distinction seems to depend on fuzzy considerations about the purpose or functional role of the concept in question. Some of the things you say also suggest that you're attempting to resuscitate a form of classical empricist sensory atomism, where the "atomic" symbols refer to sensory categories acquired "by acquaintance" and the meaning of complex symbols is built up from the atoms "by description". This approach has an honorable history in philsophy; unfortunately, no one has ever been able to make it work. In addition to the above considerations, the main problems seem to be: first, that no principled distinction can be made between the simple sensory concepts and the complex "theoretical" ones; and second, that very little that is interesting can be explicitly defined in sensory terms (try, for example, "chair"). I realize the above considerations may not be relevant to your program -- I just can't tell to what extent you expect it to shed any light on the problem of explaining semantic content in naturalistic terms. In any case, I think it's important to understand why this fundamental problem remains largely untouched by such theories. Anders Weinstein BBN Labs ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Sat Jun 27 03:30:44 1987 Date: Sat, 27 Jun 87 03:30:34 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #155 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Sat, 27 Jun 87 03:20 EDT Received: from relay.cs.net by RELAY.CS.NET id af28970; 26 Jun 87 20:02 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa08353; 26 Jun 87 20:01 EDT Date: Fri 26 Jun 1987 00:08-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #155 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Friday, 26 Jun 1987 Volume 5 : Issue 155 Today's Topics: Seminars - Acquiring Knowledge from the Outside (Rutgers) & AI Research at Edinburgh (SRI) & Nonmonotonic Multiple Inheritance Systems (Bell Labs) Conference - Advanced Computing Symposium & European Conference on AI in Medicine ---------------------------------------------------------------------- Date: 18 Jun 87 11:47:53 EDT From: KALANTARI@RED.RUTGERS.EDU Subject: Seminar - Acquiring Knowledge from the Outside (Rutgers) R U T G E R S U N I V E R S I T Y Department of Computer Science C O L L O Q U I U M SPEAKER: Paul Rosenbloom Stanford University TITLE: ACQUIRING KNOWLEDGE FROM THE OUTSIDE SOME RECENT PROGRESS ON LEARNING IN SOAR DATE: Monday, June 29, 1987 TIME: 10:00 a.m. PLACE: Hill Center, Room 705 In previous work on learning in Soar we have focused on how the chunking of internal problem solving can acquire the varieties of knowledge required by a general problem solver; for example, productions can be acquired which perform operator retrieval, instantiation, selection, and implementation. One major form of learning not covered by this previous work is the acquisition of knowledge from external sources. In this talk I will describe two current projects which are examining how the techniques utilized in the previous work can be employed to learn from external knowledge sources. The first project is working on the acquisition of general search control knowledge from external advice. This work touches on issues of operationalization, learning apprentices, analogy, and generalization. The second project is working on the acquisition of declarative knowledge. This work demonstrates for the first time in Soar what Dietterich termed "knowledge level learning"; that is, the acquisition of knowledge not already in the system's deductive closure. One implication of this demonstration is that explanation-based learning mechanisms are not inherently limited to symbol level learning. Issues that have arisen during this work include: how to decouple new facts from the context in which they were learned, how to be able to distinguish what has been learned from what hasn't, and how to index declarative information for appropriate retrieval. ------------------------------ Date: Wed, 24 Jun 87 11:32:42 PDT From: Amy Lansky Subject: Seminar - AI Research at Edinburgh (SRI) AI RESEARCH AT EDINBURGH - PAST, PRESENT, AND FUTURE Roberto Desimone University of Edinburgh 2:00 PM, FRIDAY, June 26 SRI International, Building E, Room EK242 This talk will comprise a review of AI research and other AI activities that have and are being pursued at Edinburgh. I will start with a short history of the early days of AI in Edinburgh in the 1960s and 1970s, the transition period in the mid to late 1970s and the revival in 1980s. Then, I will stress the basic research currently being conducted within the Dept. of AI at the University of Edinburgh. Some of the activities conducted within the AI Applications Institute also in Edinburgh will also be discussed. Finally, some thoughts on the future of AI research at Edinburgh. VISITORS: Please arrive 5 minutes early so that you can be escorted up from the E-building receptionist's desk. Thanks! NOTE: Different time and place ------------------------------ Date: Thu 25 Jun 1987 14:10:08 From: dlm@allegra.csnet Subject: Seminar - Nonmonotonic Multiple Inheritance Systems (Bell Labs) Date: Thursday, July 2 Time: 2:00 PM Place: AT&T Bell Laboratories MH 3D-473 David S. Touretzky Computer Science Department Carnegie Mellon University A Clash of Intuitions: The Current State of Nonmonotonic Multiple Inheritance Systems Early attempts at combining multiple inheritance with nonmonotonic reasoning were based on straightforward extensions to tree-structured inheritance systems, and were theoretically unsound. In The Mathematics of Inheritance Systems, or TMOIS, I described two basic problems that these systems cannot handle. One involves reasoning with true but redundant assertions; the other involves ambiguity. TMOIS provided the definition and analysis of a theoretically sound multiple inheritance system, accompanied by inference algorithms. Other definitions for inheritance have since been proposed by Sandewall and by Horty, Thomason, and Touretzky that are equally sound and intuitive, but do not always agree with the system defined in TMOIS. At the heart of the controversy is a clash of intuitions about certain fundamental issues such as skepticism versus credulity, the direction in which inheritance paths are extended, and classical versus intuitive notions of consistency. In this talk I will catalog the issues, map out a design space, and describe interesting properties that result from certain choices of definitions. Just as there are alternative logics, there may be no single ``best'' approach to nonmonotonic multiple inheritance. This is joint work with Richmond Thomason of the University of Pittsburgh and John Horty of CMU. Sponsor: Ron Brachman ------------------------------ Date: Thu, 18 Jun 87 13:14:49 pdt From: Douglas Schuler Subject: Conference - Advanced Computing Symposium DIRECTIONS AND IMPLICATIONS OF ADVANCED COMPUTING A One Day Symposium - July 12, 1987 University of Washington, Seattle, Washington PROGRAM On-Site Registration (8:00 - 9:00) PLENARY SESSION (9:00 - 10:30) Robert Kahn and Terry Winograd with Gary Chapman The featured speakers will discuss the role of funding on computer science research. How and why are projects selected for funding? What are the roles of the Department of Defense, civilian agencies and private sources? Does it matter where research money comes from? Robert Kahn is the founder of the non-profit Corporation for National Research Initiatives, in Washington, D.C. Until 1985, Kahn was director of the Information Processing Techniques Office at the Defense Advanced Research Projects Agency (DARPA). Terry Winograd is an associate professor of computer science at Stanford University. He is author of "Understanding Natural Language", "Language as a Cognitive Process" and (with Fernando Flores) "Understanding Computers and Cognition". Winograd is the national president of Computer Professionals for Social Responsibility (CPSR). The discussion will be moderated by Gary Chapman, Executive Director of CPSR. He is co-editor of the book, "Computers in Battle" to be published this fall. Chapman is a former member U.S. Special Forces. PARALLEL SESSIONS FUNDING (11:00 - 12:00) David Bushnell - The Promise and Reality of ARPANET: A Brief History Joel Yudken and Barbara Simons - Project on Funding in Computer Science: A Preliminary Report AI PROSPECTS I (11:00 - 12:00) Juergen Koenemann Artificial Intelligence and the Future of Work Reinhard Keil-Slawik An Ecological Approach to Responsible Systems Development LUNCH (12:00 - 1:30) MILITARY/RELIABILITY (1:30 - 3:00) Richard Hamlet - Testing for Trustworthiness David Bella - Fault-tolerant Ballistic Missile Defense Erik Nilsson - The Costs of Computing Star Wars EXPERT SYSTEMS (1:30 - 3:00) Matthew Lewis and Seth Chaikin - Will There Be Teachers in the Classroom of the Future? Rolf Engelbrecht - Expert Systems in Medicine - A Technology Assessment Carole Hafner and Donald Berman - The Potential of AI to Help Solve the Crisis in Our Legal System BREAK (3:00 - 3:30) RESEARCH PRIORITIES (3:30 - 4:30) Douglas Schuler - A Civilian Computing Initiative: Three Modest Proposals Jack Beusmans and Karen Wieckert - Artificial Intelligence and the Military AI PROSPECTS II (3:30 - 4:30) Susan Landau - The Responsible Use of 'Expert' Systems K. Eric Drexler - Technologies of Danger and Wisdom VIDEO Daressa Computers in Context CPSR Reliability and Risk videotape on DBNET (a computer mail network for the deaf-blind) Registration fees Regular $50 ____ CPSR Member $30 ____ Student/Low Income $20 ____ Proceedings only (cannot attend symposium) $15 ____ Proceedings will be distributed to symposium registrants on day of symposium. Lunch is included. DIAC '87 CPSR/Seattle P.O. Box 85481 Seattle, WA 98105 Sponsored by Computer Professionals for Social Responsibility ------------------------------ Date: 23 Jun 1987 12:32:28 EST From: Herve.Lambert@PS3.CS.CMU.EDU Subject: Conference - European Conference on AI in Medicine EUROPEAN CONFERENCE on ARTIFICIAL INTELLIGENCE in MEDICINE __________ Marseilles (France), Aug 31st - Sept 3rd Organized by: AIME, European Society for Artificial Intelligence in Medicine In cooperation with: IIRIAM, International Institute of Robotics and Artificial Intelligence of Marseilles. IRCF, Imperial Cancer Research Fund Laboratories, UK GSF-MEDIS, Gesellschaft fur Strahlen und umweltforschung mbH Munchen Laboratoire d'Informatique Medicale de la Faculte de Medecine de Marseille. PROGRAM _______ WORKSHOP and TUTORIALS ---------------------- Monday, August 31st: Tutorial 1: 9.00 - 13.00 Acquisition of Knowledge from Medical databases Gio C.M. Wiederhold, M. Walker, R.L. Blum, Stanford University (USA) Tutorial 2: 14.00 - 18.00 Methods and Techniques used in Expert Systems Jan L. Talmon, Henny P.A. Boshuizen, University of Limburg (NL) Tutorial 3: 14.00 - 18.00 Knowledge representation Steen Andreassen, University of Aalborg (DK), Mike Wellman, MIT (USA) Workshop: 9.00 - 13.00 From Mycin to Oncocin Larry Fagan, Stanford University (USA) CONFERENCE ---------- Tuesday, September 1st, 1987 9.00 - 9.30 Opening Session 9.30 - 10.30 Invited Keynote speaker: J.H. Van Bemmel, Free University of Amsterdam 10.30 - 11.00 Break Session 1: Methodology 11.00 - 11.30 "INTERMED": a medical language interface. Mery C., Normier B., Orgonowski A. (F) 11.30 - 12.00 Inference engineering through prototyping in Prolog Van Thilo J., Mulders A. (B) 12.00 - 12.30 The evaluation of clinical decision support systems: a discussion of the methodology used in the ACORN project. Wyatt J. (UK) 12.30 - 13.00 Matching patients: an approach to decision support in Liver transplantation. Tusch G., Bernauer J., Reichertz P.L. (FRG) 13.00 - 14.00 Lunch Session 2: Clinical Applications (1) 14.30 - 15.00 An expert system for diagnosis and therapy planning in patients with peripheral vascular disease. Talmon J.L., Schijven R.A.J., Kitslaar P.J.E.H.M., Penders R. (NL) 15.00 - 15.30 An expert system for the classification of Dizziness and Vertigo. Schmid R., Zanocco P., Buizza A., Magenes G., Manfrin M., Mira E. (I) 15.30 - 16.00 The SENEX system, a microcomputer-based expert system built by oncologists for breast cancer management. Renaud-Salis J.L., Bonichon F., Durand M., Avril A., Lagarde C. (F). 16.00 - 16.30 Break Session 3: Qualitative Reasoning 16.30 - 17.00 The use of QSIM for Qualitative simulation of physiological systems. Nicolosi E., Leaning M. (UK) 17.00 - 17.30 Qualitative description of electrophysiologic measurements: toward automatic data interpretation. Irler W.J., Antolini R., Kirchner M., Stringa L. (I) 17.30 - 18.00 A qualitative spatial representation for cardiac Electrophysiology. Gotts N. (UK) 18.45 Cocktail at the city hall of Marseilles. Wednesday September 2nd 1987{ Session 4: Knowledge acquisition and representation 9.00 - 9.30 Knowledge acquisition in expert system assisted diagnosis: a machine learning approach Funk M., Appel R.D., Roch Ch., Hochstrasse D., Pellegrini Ch., Muller A.F., (CH) 9.30 - 10.00 Knowledge representation for cooperative medical systems Rector A.L. (UK) 10.00 - 10.30 A representation of time for medical expert systems Hamlet I., Hunter J. (UK) 10.30 - 11.00 Break Session 5: Management of uncertainty 11.00 - 11.30 TOULMED an inference engine which deals with imprecise and uncertain aspects of medical knowledge. Buisson J.C., Farreny H., Prade H., Turnin M.C., Tauber J.P., Bayard F. (F) 11.30 - 12.00 Coherent handling of uncertainty via localized computation in an expert system for therapeutic decision. Berzuini C., Barosi G., Polino G. (I) 12.00 - 12.30 MUNIN, on the case for probabilities in medical expert systems a pratical exercise. Jensen F.V., Andersen S.K., Kjaerulff U., Andreassen S. (DK) 12.30 - 13.00 Rule based expert systems in gynecology: statistical versus heuristic approach Riss P.A., Koelbl H., Reinthaller A., Deutinger J. (Austria) Afternoon: Social Program Thursday September 3rd Session 6: Knowledge Engineering tools. 9.00 - 9.30 A radiological expert system for the P.C., design and Implementation issues. Horn W., Imhof H., Pfahringer B., Salamonowitz E., (Austria) 9.30 - 10.00 A P.C based shell for clinical information systems with reasoning capabilities Wiener F., Groth T. (Israel, Sweden) 10.00 - 10.30 The kernel mechanism for handling assumptions and justifications and its applications to the biotechnologies Cherubini M.A., Cerri S.A., Sbarbati R. (I) 10.30 - 11.00 Break Session 7: General Session 11.00 - 12.00 Invited Lecture Larry Fagan, Stanford University (USA) 12.00 - 12.30 Man-machine interaction in CHECK Console L., Fossa M., Torasso P., Molino G., Cravetto G (I) 12.30 - 13.00 The Oxford system of medicine Fox J., Glowinski A., O'Neil M. (UK) 13.00 - 14.30 Lunch Session 8: Clinical Applications (2) 14.30 - 15.00 Evaluating the performance of AMEMIA Quaglini S., Stefanelli M., Barosi G., Berzuini A. (I) 15.00 - 15.30 Computer aided diagnosis and treatment of brachial plexus injuries. Jaspers R.B.M., Van der Helm F.C.T. (NL) 15.30 - 16.00 Representation of embryonic development and its anomalies. Goutal J.M., Philip N., Griffiths M., Ayme S. (F) 16.00 - 16.30 A microcomputer based decision support for lipid desorder. Fhaircheallaigh D.N., Sinnot M., Grimson J., O'Moore R. (Eire) 16.30 - 17.00 Closing session Program Committee: J. Fox, London Chairman P. Adlassnig, Vienna R. Engelbrecht, Munich Tutorial M. Fieschi, Marseille F. Gremy, Montpellier (F) T. Groth, Upsalla A. Hasman, Maastricht A.L., Rector, Manchester P.L. reichertz, Hannover P. Smets, Brussels M. Stefanelli, Pavia Organizing committee: M. Fieschi Chairman V. Bernadac Organization P. Dujol B. Guisiano Social events M. Joubert Local arrangements D. Riouall Liaison M. Roux G. Soula Exhibition Additionnal Informations: Viviane Bernadac IIRIAM 2 rue Henri Barbusse, CMCI 13241 Marseille Cedex 1 - FRANCE tel: (33) 91 91 36 72 telex: 440 860 telefax: (33) 91 91 70 24 ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 30 08:20:57 1987 Date: Tue, 30 Jun 87 08:20:52 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #156 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 30 Jun 87 07:51 EDT Received: from relay.cs.net by RELAY.CS.NET id aa11054; 29 Jun 87 2:15 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa22196; 29 Jun 87 2:14 EDT Date: Sun 28 Jun 1987 22:22-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #156 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 29 Jun 1987 Volume 5 : Issue 156 Today's Topics: Theory - Symbol Grounding and Invertibility ---------------------------------------------------------------------- Date: Mon, 22 Jun 87 10:19:59 PDT From: Neil Hunt Subject: Symbol grounding and invertibility. John Cugini writes: > What if there were a few-to-one transformation between the skin-level > sensors ... > My example was to suppose that #1: > a combination of both red and green retinal receptors and #2 a yellow > receptor BOTH generated the same iconic yellow. We humans see the world (to a first order at least) through red, green and blue receptors. We are thus unable to distinguish between light of a yellow frequency, and a mixture of light of red and green frequencies, and we assign to them a single token - yellow. However, if our visual apparatus was equipped with yellow receptors as well, then these two input stimuli would *appear* quite different, as indeed they are. In this case I think that it is highly unlikely that we would have the same symbol to represent the two cases. Consider a species with only two classes of colour receptors, low frequency and high frequency, roughly equivalent to our concepts of red and blue, but with no middle frequency receptors corresponding to a human concept of green). Creatures of such a species when shown pure green light would receive reduced levels from the receptors on each side of green frequency, thus receiving some combination of blue and red signals. This would be indistinguishable from a mixture of blue and red, which we call magenta. Such creatures might then reason (incorrectly) about the possibility of having a middle frequency receptor, and having a many to one mapping between case #1, pure green light, and case #2, a mixture of red and blue, and wonder about how that affects questions of invertibility. As we humans know, if these creatures had such a visual capability, they would invent a new symbol for magenta, and there would be no many to one mapping. > Clearly this iconic representation is non-invertible back out to the > sensory surfaces, but intuitively it seems like it would be grounded > nonetheless - how about it? The fallacy is that iconic representation described is indeed non invertible, but it is also clearly not grounded, since if we had yellow receptors, we would be able to perceive a difference between, and require a new symbol for one of the new colours. Neil/. ----- End Forwarded Message ----- ------------------------------ Date: 21 Jun 87 22:55:09 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <6670@diamond.BBN.COM>, aweinste@Diamond.BBN.COM (Anders Weinstein) writes, with reference to article <861@mind.UUCP> harnad@mind.UUCP (Stevan Harnad): > > Some of the things you say also suggest that you're attempting to resuscitate > a form of classical empricist sensory atomism, where the "atomic" symbols > refer to sensory categories acquired "by acquaintance" and the meaning of > complex symbols is built up from the atoms "by description". This approach > has an honorable history in philsophy; unfortunately, no one has ever been > able to make it work. In addition to the above considerations, the main > problems seem to be: first, that no principled distinction can be made > between the simple sensory concepts and the complex "theoretical" ones; and > second, that very little that is interesting can be explicitly defined in > sensory terms (try, for example, "chair"). > I hope none of us are really trying to resuscitate classical philosophies, because the object of this discussion is to learn how to use modern technologies. To define an interesting object in sensory terms requires an intermediary module between the sensory system and the symbolic system. With a chair in the visual sensory field, the system will use hard-coded nonlinear (decision-making) techniques to identify boundaries and shapes of objects, and identify the properties that are invariant to rotation and translation. A plain wooden chair and an overstuffed chair will be different objects in these terms. But the system might also learn to identify certain types of objects that move, i.e., those we call people. If it notices that people assume the same position in association with both chair-objects, it could decide to use the same category for both. The key to this kind of classification is that the chair is not defined in explicit sensory terms but in terms of filtered sensory input. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 P.S. Sorry for the double posting of my previous article. ------------------------------ Date: 20 Jun 87 02:17:09 GMT From: ihnp4!homxb!houxm!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <861@mind.UUCP>, harnad@mind.UUCP writes: > marty1@houdi.UUCP (M.BRILLIANT) asks: > > > what do you think is essential: (A) literally analog transformation, > > (B) invertibility, or (C) preservation of significant relational > > functions? > Let me see if I can correctly rephrase his answer: (i) "discrimination" (pairwise same/different judgments) he associates with iconic ("analog") representations, which he says have to be invertible, and will ordinarily be really analog because "dedicated" digital equivalents will be too complex. (ii) for "identification" or "categorization" (sorting and labeling of objects), he says only distinctive features need be extracted from the sensory projection; this process is not invertible. (iii) for "conscious problem-solving," etc., he says relation-preserving symbolic representations would be optimal, if they are not "autonomous (modular)" but rather are grounded by deriving their atomic symbols through the categorization process above. (iv) to pass the Total Turing Test he wants all of the above, tied together in the sequence described. I agree with this formulation in most of its terms. But some of the terms are confusing, in that if I accept what I think are good definitions, I don't entirely agree with the statements above. "Invertible/Analog": The property of invertibility is easy to visualize for continuous functions. First, continuous functions are what I would call "analog" transformations. They are at least locally image-forming (iconic). Then, saying a continuous transformation is invertible, or one-to-one, means it is monotonic, like a linear transformation, rather than many-to-one like a parabolic transformation. That is, it is unambiguously iconic. It might be argued that physical sensors can be ambiguously iconic, e.g., an object seen in a half-silvered mirror. Harnad would argue that the ambiguity is inherent in the physical scene, and is not dependent on the sensor. I would agree with that if no human sensory system ever gave ambiguous imaging of unambiguous objects. What about the ambiguity of stereophonic location of sound sources? In that case the imaging (i) is unambiguous; only the perception (ii) is ambiguous. But physical sensors are also noisy. In mathematical terms, that noise could be modeled as discontinuity, as many-to-one, as one-to-many, or combinations of these. The noisy transformation is not invertible. But a "physically analog" sensory process (as distinct from a digital one) can be approximately modeled (to within the noise) by a continuous transformation. The continuous approximation allows us to regard the analog transformation as image-forming (iconic). But only the continuous approximation is invertible. "Autonomous/Modular": The definition of "modular" is not clear to me. I have Harnad's definition "not analogous to a top-down, autonomous symbol-crunching module ... hardwired to peripheral modules." The terms in the definition need defining themselves, and I think there are too many of them. I would rather look at the "hybrid" three-layer system and say it does not have a "symbol-cruncher hardwired to peripheral modules" because there is a feature extractor (and classifier) in between. The main point is the presence or absence of the feature extractor. The symbol-grounding problem arises because the symbols are discrete, and therefore have to be associated with discrete objects or classes. Without the feature extractor, there would be no way to derive discrete objects from the sensory inputs. The feature extractor obviates the symbol-grounding problem. I consider the "symbol-cruncher hardwired to peripheral modules" to be not only a straw man but a dead horse. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 26 Jun 87 04:38:02 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem John Cugini on ailist@stripe.sri.com writes: > What if there were a few-to-one transformation between the skin-level > sensors (remember Harnad proposes "skin-and-in" invertibility > as being necessary for grounding) and the (somewhat more internal) > iconic representation. My example was to suppose that #1: > a combination of both red and green retinal receptors and #2 a yellow > receptor BOTH generated the same iconic yellow. > Clearly this iconic representation is non-invertible back out to the > sensory surfaces, but intuitively it seems like it would be grounded > nonetheless - how about it? Invertibility is a necessary condition for iconic representation, not for grounding. Grounding symbolic representations (according to my hypothesis) requires both iconic and categorical representations. The latter are selective, many-to-few, invertible only in the features they pick out and, most important, APPROXIMATE (e.g., as between red-green and yellow in your example above). This point has by now come up several times... -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 26 Jun 87 05:07:40 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem: McCarthy's query In article 208 of comp.ai.digest: JMC@SAIL.STANFORD.EDU (John McCarthy) asks: > I imagine that the alleged point at issue and a few of the positions > taken could be summarized for the benefit of those of us whose > subjective probability that there is a real point at issue is too > low to motivate studying the entire discussion but high enough to > motivate reading a summary. The point at issue concerns how symbols in a symbol-manipulative approach to the modeling of mind can be grounded in something other than more symbols so that their meanings and their connections to objects can be independent of people's interpretations of them. One of the positions taken was that connecting a purely symbolic module to peripheral (transducer/effector) modules in the right way should be all you need to ground the symbols. I suggested that all this is likely to yield is more of the toy models that symbolic AI has produced until now. To get human-scale (Total Turing Test) performance capacity, a bottom-up hybrid nonsymbolic/symbolic system may be needed, one in which the elementary symbols are the names of sensory categories picked out by inductive (possibly connectionist) feature-filters (categorical representations) and invertible analogs of sensory projections (iconic representations). This model is described in "Categorical Perception: The Groundwork of Cognition" (Cambridge University Press 1987, S. Harnad, ed., ISBN 0-521-26758-7). Other alternatives that have been mentioned by others in the discussion included: (1) symbol-symbol "grounding" is already enough and (2) connectionist nets already generate grounded "symbols." If you want the entire file, I've saved it all... -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 26 Jun 87 17:19:29 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <914@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > Invertibility is a necessary condition for iconic representation, not > for grounding. Grounding symbolic representations (according to my > hypothesis) requires both iconic and categorical representations... Syllogism: (a) grounding ... requires ... iconic ... representation.... (b) invertibility is ... necessary ... for iconic representation. (c) hence, grounding must require invertibility. Why then does harnad say "invertibility is a necessary condition for ..., NOT for grounding" (caps mine, of course)? This discussion is getting hard to follow. Does it have to be carried on simultaneously in both comp.ai and comp.cog-eng? Could harnad, who seems to be the major participant, pick one? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 26 Jun 87 18:03:26 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem: McCarthy's query Will the proponents of the various views described below, and those whose revelant views have not been described below, please stand up? In article <915@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > In article 208 of comp.ai.digest: JMC@SAIL.STANFORD.EDU (John McCarthy) > asks: > > > I imagine that the alleged point at issue and a few of the positions > > taken could be summarized ..... > > The point at issue concerns how symbols in a symbol-manipulative > approach to the modeling of mind can be grounded in something other > than more symbols so that their meanings and their connections to > objects can be independent of people's interpretations of them. > ..... One of > the positions taken was that connecting a purely symbolic module to > peripheral (transducer/effector) modules IN THE RIGHT WAY should be > all you need to ground the symbols. Caps mine. Position 1 is that the peripherals and the symbolic module have to be connected in the right way. Harnad's position is that > .... a bottom-up hybrid nonsymbolic/symbolic system may be > needed, one in which the elementary symbols are the names of sensory > categories picked out by inductive (possibly connectionist) feature-filters > (categorical representations) and invertible analogs of sensory projections > (iconic representations)..... This looks like a way to connect periperals to a symbolic module. To the extent that I understand it, I like it, except for the invertibility condition. If it's the right way, it's a special case of position 1. Harnad has called the "right way" of position 1 "top-down," "hard-wired," and other names, to distance himself from it. I'm not sure there are any real proponents of position 1 in such a narrow sense. I support position 1 in the wide sense, and I think Harnad does. > ..... Other alternatives that have been > mentioned by others in the discussion included: (1) symbol-symbol "grounding" > is already enough .... They don't care about the problem, so either they or we can go away. They (and I) want this discussion to go to another newsgroup. > ..... and (2) connectionist nets already generate grounded "symbols." Is that a variant of Harnad's position, i.e., "(possibly connectionist)"? I think the real subject of discussion is the definition of some of the technical terms in Harnad's position, and the identification of which elements are critical and which might be optional? Might some of the disagreement disappear if the definitions were more concrete? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Mon Jun 29 03:31:04 1987 Date: Mon, 29 Jun 87 03:30:56 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #157 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Mon, 29 Jun 87 03:21 EDT Received: from relay.cs.net by RELAY.CS.NET id aa11103; 29 Jun 87 2:28 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa22235; 29 Jun 87 2:25 EDT Date: Sun 28 Jun 1987 22:26-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #157 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 29 Jun 1987 Volume 5 : Issue 157 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 26 Jun 87 19:41:11 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem berleant@ut-sally.UUCP (Dan Berleant) of U. Texas CS Dept., Austin, Texas writes: > Are you saying that the categorical representations are to be > nonsymbolic? The review of human concept representation I recently read > (Smith and Medin, Categories and Concepts, 1981) came down... hard on > the holistic theory of concept representation... The alternative > nonsymbolic approach would be the 'dimensional' one. It seems a > strongish statement to say that this would be sufficient, to the > exclusion of symbolic properties... However, the metric > hypothesis -- that a concept is sufficiently characterized by a point > in a multi-dimensional space -- seems wrong, as experiments have shown. Categorical representations are the representations of purely SENSORY categories, and I am indeed saying that they are to be NONsymbolic. Let me also point out that the theory I am putting forward represents a direct challenge to the Roschian line of category research in which the book you cite belongs. To put it very briefly, I claim that that line of experimental and theoretical work is not really investigating the representations underlying the capacity to categorize at all; it is only looking at the fine tuning of category judgments. The experiments are typically not addressing the question of how it is that a device or organism can successfully categorize the inputs in question in the first place; instead it examines (1) how QUICKLY or EASILY subjects do it, (2) how TYPICAL (of the members of the category in question) subjects rate the inputs to be and (3) what features subjects INTROSPECT that they are using. This completely bypasses the real question of how anyone or anything actually manages to accomplish the categorization at all. Let me quickly add that there is nothing wrong with reaction-time experiments if they suggest hypotheses about the basic underlying mechanism, or provide ways of testing them. But in this case -- as in many others in experimental cognitive psychology -- the basic mechanisms are bypassed and the focus is on fine-tuning questions that are beside the point (or premature) -- if, that is, the objective is to explain how organisms or devices actually manage to generate successful categorization performance given the inputs in question. As an exercise, see where the constructs you mention above -- "holistic," "dimensional," or "metric" representations -- are likely to get you if you're actually trying to get a device to categorize, as we do. There is also an "entry point" problem with this line of research, which typically looks willy-nilly at higher-order, abstract categories, as well as "basic level" object categories (an incoherent concept, in my opinion, except as an arbitrary default level), and even some sensory categories. But it seems obvious that the question of how the higher-order categories are represented is dependent on how the lower-order ones are represented, the abstract ones on the concrete ones, and perhaps all of these depend on the sensory ones. Moreover, often the inputs used are members of familiar, overlearned categories, and the task is a trivial one, not engaging the mechanisms that were involved in their acquisition. In other experiments, artificial stimuli are used, but it is not clear how representative these are of the category acquisition process either. Finally, and perhaps most important: In bypassing the problem of categorization capacity itself -- i.e., the problem of how devices manage to categorize as correctly and successfully as they do, given the inputs they have encountered -- in favor of its fine tuning, this line of research has unhelpfully blurred the distinction between the following: (a) the many all-or-none categories that are the real burden for an explanatory theory of categorization (a penguin, after all, be it ever so atypical a bird, and be it ever so time-consuming for us to judge that it is indeed a bird, is, after all, indeed a bird, and we know it, and can say so, with 100% accuracy every time, irrespective of whether we can successfully introspect what features we are using to say so) and (b) true "graded" categories such as "big," "intelligent," etc. Let's face the all-or-none problem before we get fancy... > To discuss "invariant features... sufficient to guide reliable > categorization" sounds like the "classical" theory (as Smith & Medin > call it) of concept representation: Concepts are represented as > necessary and sufficient features (i.e., there are defining features, > i.e. there is a boolean conjunction of predicates for a concept). This > approach has serious problems, not the least of which is the inability > of humans to describe these features for seemingly elementary concepts, > like "chair", as Weinstein and others point out. I contend that a > boolean function (including ORs as well as ANDs) could work, but that > is not what was mentioned. An example might be helpful: A vehicle must > have a steering wheel OR handlebars. But to remove the OR by saying, > a vehicle must have a means of steering, is to rely on a feature which > is symbolic, high level, functional, which I gather we are not allowing. It certainly is the "classical" theory, but the one with the serious problems is the fine-tuning approach I just described, not the quite reasonable assumption that if 100% correct, all-or-none categorization is possible at all (without magic), then there must be a set of features in the inputs that is SUFFICIENT to generate it. I of course agree that disjunctive features are legitimate -- but whoever said they weren't? That was another red herring introduced by this line of research. And, as I mentioned, "the inability of humans to describe these features" is irrelevant. If they could do it, they'd be cognitive modelers! We must INFER what features they're using to categorize successfully; nothing guarantees they can tell us. (If by "Weinstein" you mean "Wittgenstein" on "games," etc., I have to remind you that Wittgenstein did not have the contemporary burden of speaking in terms of internal mechanisms a device would have to have in order to categorize successfully. Otherwise he would have had to admit that "games" are either (i) an all-or-none category, i.e., there is a "right" or "wrong" of the matter, and we are able to sort accordingly, whether or not we can introspect the basis of our correct sorting, or (ii) "games" are truly a fuzzy category, in which membership is arbitrary, uncertain, or a matter of degree. But if the latter, then games are simply not representative of the garden-variety all-or-none categorization capacity that we exercise when we categorize most objects, such as chairs, tables, birds. And again, there's nothing whatsoever wrong with disjunctive features.) Finally, it is not that we are not "allowing" higher-order symbolically described features. They are the goal of the whole grounding project. But the approach I am advocating requires that symbolic descriptions be composed of primitive symbols which are in turn the labels of sensory categories, grounded in nonsymbolic (iconic and categorical) representations. > [Concerning model-theoretic "grounding":] The more statements > you have (that you wish to be deemed correct), the more the possible > meanings of the terms will be constrained. To illustrate, consider > the statement FISH SWIM. Think of the terms FISH and SWIM as variables > with no predetermined meaning -- so that FISH SWIM is just another way > of writing A B. What variable bindings satisfy this? Well, many do... > Now consider the statement FISH LIVE, where FISH and LIVE are variables. > Now there are two statements to be satisfied. The assignment to the > variable LIVE restricts the possible assignments to the variable SWIM... > Of course, we have many many statements in our minds that must be > simultaneously satisfied, so the possible meanings that each word name > can be assigned is correspondingly restricted. Could the restrictions be > sufficient to require such a small amount of ambiguity that the word > names could be said to have intrinsic meaning?... footnote: This > leaves unanswered the question of how the meanings themselves are > grounded. Non-symbolically, seems to be the gist of the discussion, > in which case logic would be useless for that task even in an > "in principle" capacity since the stuff of logic is symbols. I agree that there are constraints on the correlations of symbols in a natural language, and that the degrees of freedom probably shrink, in a sense, as the text grows. That is probably the basis of successful cryptography. But I still think (and you appear to agree) that even if the degrees of freedom are close to zero for a natural language's symbol combinatons and their interpretations, this still leaves the grounding problem intact: How are the symbols connected to their referents? And what justifies our interpretation of their meanings? With true cryptography, the decryption of the symbols of the unknown language is always grounded in the meanings of the symbols of a known language, which are in turn grounded in our heads, and their understanding of the symbols and their relation to the world. But that's the standard DERIVED meaning scenario, and for cognitive modeling we need INTRINSICALLY grounded symbols. (I do believe, though, that the degrees-of-freedom constraint on symbol combinations does cut somewhat into Quine's claims about the indeterminacy of radical translation, and ESPECIALLY for an intrinsically grounded symbol system.) -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 26 Jun 87 22:17:16 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem aweinste@Diamond.BBN.COM (Anders Weinstein) of BBN Laboratories, Inc., Cambridge, MA writes: > I don't see any difference between "physical" and "merely theoretical" > invertibility... Surely you don't mean that a transformation-inversion > capability must actually be present in the device for it to count as > "analog" in your sense. (Else brains, for example, wouldn't count). I think this is partly an empirical question. "Physically possible" invertibility is enough for an analog transformation, but actual physical invertibility may be necessary for an iconic representation that can generate all of our discrimination capacities. Avoiding "merely theoretical" invertibility is also part of avoiding any reliance on mediation by our theoretical interpretations in order to get an autonomous, intrinsically grounded system. > the *semantic* meaning of a symbol is still left largely unconstrained > even after you take account of it's "grounding" in perceptual > categorization. This is because what matters for intentional content > is not the objective property in the world that's being detected, but > rather how the subject *conceives* of that external property, a far > more slippery notion... primitive people may be able to reliably > categorize certain large-scale atmospheric electrical discharges; > nevertheless, the semantic content of their corresponding states might > be "Angry gods nearby" or some such. I agree that symbol grounding cannot be based on the "objective property" that's being detected. Categorical representations in my grounding model are approximate. All they do is sort and label the confusable alternatives that have been sampled, using the provisional features that suffice to generate reliable sorting performance according to the feedback that defines "right" and "wrong." There is always a context of confusable alternatives, and which features are used to sort reliably is always a "compared to what?" matter. The exact "objective property" they pick out is never an issue, only whether they can generate reliable asymptotic categorization performance given that sample and those feedback constraints. The representation is indifferent to whether what you are calling "water," is really "twin-water" (with other objective properties), as long as you can sort it "correctly" according to the feedback (say, from the dictates of thirst, or a community of categorizing instructors). As to what people "conceive" themselves to be categorizing: My model is proposed in a framework of methodological epiphenomenalism. I'm interested in what's going on in people's heads only inasmuch as it is REALLY generating their performance, not just because they think or feel it is. So, for example, in criticizing the Roschian approach to categorization in my reply to Dan Berleant I suggested that it was irrelevant what features subjects BELIEVED they were using to categorize, say, chairs; what matters is what features they (or any organism or device in a similar input situation) really ARE using. [This does not contradict my previous point about the irrelevance of "objective properties." "Features" refers to properties of the proximal projection on the device's sense receptors, whereas "properties" would be the essential characteristics of distal objects in the world. Feature detectors are blind to distal differences that are not preserved in the proximal projection.] On the other hand, "Angry gods nearby" is not just an atomic label for "thunder" (otherwise it WOULD be equivalent to it in my model -- both labels would pick out approximately the same thing); in fact, it is decomposable, and hence has a different meaning in virtue of the meanings of "angry" and "gods." There should be corresponding internal representational differences (iconic, categorical and symbolic) that capture that difference. > Another well-known obstacle to moving from an objective to an > intentional description is that the latter contains an essentially > normative component, in that we must make some distinction between > correct and erroneous classification. For example, we'd probably > like to say that a frog has a fly-detector which is sometimes wrong, > rather than a "moving-spot-against-a- fixed-background" detector > which is infallible. Again, this distinction seems to depend on fuzzy > considerations about the purpose or functional role of the concept > in question... [In his reply on this point to Dan Berleant, > Weinstein continues:] the philosophical problem is to say why any > response should count as an *error* at all. What makes it wrong? > I.e. who decides which "concept" -- "fly" or "moving-spot..." -- the > frog is trying to apply? The objective facts about the frog's > perceptual abilities by themselves don't seem to tell you that in > snapping out its tongue at a decoy, it's making a *mistake*. To > say this, an outside interpreter has to make some judgement about what > the frog's brain is trying to accomplish by its detection of moving > spots. And this makes the determination of semantic descriptions a > fuzzy matter. I don't think there's any problem at all of what should count as an "error" for my kind of model. The correctness or incorrectness of a label is always determined by feedback -- either ecological, as in evolution and daily nonverbal learning, or linguistic, where it is conventions of usage that determine what we call what. I don't see anything fuzzy about such a functional framework. (The frog's feedback, by the way, probably has to do with edibility, so (i) "something that affords eating" is probably a better "interpretation" of what it's detecting. And, to the extent that (ii) flies and (iii) moving spots are treated indifferently by the detector, the representation is approximate among all three. The case is not like that of natives and thunder, since the frog's "descriptions" are hardly decomposable. Finally, there is again no hope of specifying distal "objective properties" ["bug"/"schmug"] here either, as approximateness continues to prevail.) > Some of the things you say also suggest that you're attempting to > resuscitate a form of classical empricist sensory atomism, where the > "atomic" symbols refer to sensory categories acquired "by acquaintance" > and the meaning of complex symbols is built up from the atoms "by > description". This approach has an honorable history in philosophy; > unfortunately, no one has ever been able to make it work. In addition > to the above considerations, the main problems seem to be: first, > (1) that no principled distinction can be made between the simple > sensory concepts and the complex "theoretical" ones; and second, > (2) that very little that is interesting can be explicitly defined in > sensory terms (try, for example, "chair")...[In reply to Berleant, > Weinstein continues:] Of course *some* concepts can be acquired by > definition. However, the "classical empiricist" doctrine is committed > to the further idea that there is some privileged set of *purely > sensory* concepts and that all non-sensory concepts can be defined in > terms of this basis. This is what has never been shown to work. If you > regard "juice" as a "primitive" concept, then you do not share the > classical doctrine. (And if you do not, I invite you try giving > necessary and sufficient conditions for juicehood.) You're absolutely right that this is a throwback to seventeenth-century bottom-upism. In fact, in the CP book I call the iconic and categorical representations the "acquaintance system" and the symbolic representations the "description system." The only difference is that I'm only claiming to be giving a theory of categorization. Whether or not this captures "meaning" depends (for me at any rate) largely on whether or not such a system can successfully pass the Total Turing Test. It's true that no one has made this approach work. But it's also true that no one has tried. It's only in today's era of computer modeling, robotics and bioengineering that these mechanisms will begin to be tested to see whether or not they can deliver the goods. To reply to your "two main problems": (1) Even an elementary sensory category such as "red" is already abstract once you get beyond the icon to the categorical representation. "Red" picks out the electromagnetic wave-lengths that share the feature of being above and below a certain threshold. That's an abstraction. And in exchange for generating a feature-detector that reliably picks it out, you get a label -- "red" -- which can now enter into symbolic descriptions (e.g., "red square"). Categorization is abstraction. As soon as you've left the realm of invertible icons, you've begun to abstract, yet you've never left the realm of the senses. And so it goes, bottom up, from there onward. (2) As to sensory "definitions": I don't think this is the right thing to look for, because it's too hard to find a valid "entry point" into the bottom-up hierarchy. I doubt that "chair" or "juice" are sensory primitives, picked out purely by sensory feature detectors. They're probably represented by symbolic descriptions such as "things you can sit on" and "things you can drink," and of course those are just the coarsest of first approximations. But the scenario looks pretty straightforward: Even though it's flexible enough to be revised to include a chair (suitably homegenized) as a juice and a juice (for a bug?) as a chair, it seems very clear that it is the resources of (grounded) symbolic description that are being drawn upon here in picking out what is and is not a chair, and on the basis of what features. The categories are too interrelated (and approximate, and provisional) for an exhaustive "definition," but provisional descriptions that will get you by in your sorting and labeling -- and, more important, are revisable and updatable, to tighten the approximation -- are certainly available and not hard to come by. "Necessary and sufficient conditions for juicehood," however, are a red herring. All we need is a provisional set of features that will reliably sort the instances as environmental and social feedback currently dictates. Remember, we're not looking for "objective properties" or ontic essences -- just something that will guide reliable sorting according to the contingencies sampled to date. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 30 08:21:23 1987 Date: Tue, 30 Jun 87 08:21:13 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #158 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 30 Jun 87 07:54 EDT Received: from relay.cs.net by RELAY.CS.NET id af11103; 29 Jun 87 2:31 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa22249; 29 Jun 87 2:28 EDT Date: Sun 28 Jun 1987 22:30-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #158 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 29 Jun 1987 Volume 5 : Issue 158 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 27 Jun 87 01:09:41 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem: McCarthy's query marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel writes: > But a "physically analog" sensory process (as distinct from a digital > one) can be approximately modeled (to within the noise) by a continuous > transformation. The continuous approximation allows us to regard the > analog transformation as image-forming (iconic). But only the > continuous approximation is invertible. I have no quarrel with this, in fact I make much the same point -- that iconic representations are approximate too -- in the chapter describing the three kinds of representation. Is there any reason for expecting I would object? > the "hybrid" three-layer system... does not have a "symbol-cruncher > hardwired to peripheral modules" because there is a feature extractor > (and classifier) in between. The main point is the presence or > absence of the feature extractor... The symbol-grounding problem > arises because the symbols are discrete, and therefore have to be > associated with discrete objects or classes. Without the feature > extractor, there would be no way to derive discrete objects from the > sensory inputs. The feature extractor obviates the symbol-grounding > problem. The problem certainly is not just that of discrete symbols needing to pick out discrete objects. You are vastly underestimating the problem of sensory categorization, sensory learning, and the relation between lower and higher-order categories. Nor is it obvious that symbol manipulation can still be regarded as just symbol manipulation when the atomic symbols are constrained to be the labels of sensory categories. That's a bottom-up constraint, and symbolic AI normally expects to float down onto its sensors top-down. Imagine if your "setq" statements were constrained by what your elementary symbols were connected to, and their respective causal interrelations with other nonsymbolic sensory representations and their associated labels. > Why does Harnad say "invertibility is a necessary condition > for iconic representations..., NOT for grounding" Because the original statement of mine that you quote was a reply to a query about whether ALL representations had to be invertible for grounding. (It was accompanied by alleged counterexamples -- grounded but noninvertible percepts.) My reply indicated that only iconic ones had to be invertible, but that both iconic and categorical (noninvertible) ones were needed to ground symbols. > Position 1 [on the symbol grounding problem] is that the peripherals > and the symbolic module have to be connected in the right way. Harnad's > position is... a special case of position 1. I'm afraid not. I don't think there will be independent peripheral modules and symbolic modules suitably interconnected in the hybrid device that passes the Total Turing Test. I think a lot of what we consider cognition will be going on in the nonsymbolic iconic and categorical systems (discrimination, categorization, sensory learning and generalization) and that symbol manipulation will be constrained in ways that don't leave it in any way analogous to the notion of an independent functional module, operating on its own terms (as in standard AI), but connected at some critical point with the nonsymbolic modules. When I spoke earlier of the "connections" of the atomic symbols I had in mind something much more complexly interdigitated and interdependent than can be captured by anything that remotely resembles position 1. Position 1 is simply AI's pious hope that a pure "top-down" approach can expect to meet up with a bottom-up one somewhere in between. Mine is not a special case of this; it's a rival. > "...and (2) connectionist nets already generate grounded "symbols." Is > that a variant of Harnad's position, i.e., "(possibly connectionist)"? No. In my model connectionistic processes are just one possible candidate for the mechanism that finds the features that will reliably pick out a learned category. They would just be a component in the categorical representational system. But there are much more ambitious connectionistic views than that, for example, that connectionism can usurp the role of symbolic representations altogether or (worse) that they ARE symbolic (in some yet to be established sense). As far as I'm concerned, the latter would entail a double grounding problem for connectionism, the first to ground its interpretation of its states as symbolic states, and then to ground the interpretations of the symbolic states themselves (which is the standard symbol grounding problem). -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 27 Jun 87 14:32:42 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem: Correction re. Approximationism In responding to Cugini and Brilliant I misinterpreted a point that the former had made and the latter reiterated. It's a point that's come up before: What if the iconic representation -- the one that's supposed to be invertible -- fails to preserve some objective property of the sensory projection? For example, what if yellow and blue at the receptor go into green at the icon? The reply is that an analog representation is only analog in what it preserves, not in what it fails to preserve. Icons are hence approximate too. If all retinal squares, irrespective of color, go into gray icons, I have icons of the squareness, but not of the colors. Or, to put it another way, the grayness is approximate as between all the actual colors (and gray). There is no requirement that all the features of the sensory projection be preserved in icons; just that some of them should be -- enough to subserve our discrimination capacities. This is analogous to the fact that the sensory projection itself need not (and does not, and cannot) preserve all of the properties of the distal object. To those it fails to preserve -- and that we cannot detect by instruments or inference -- we are fated to remain "blind." But none of this information loss in either sensory projections or icons (or, for that matter, categorical representations) compromises groundedness. It just means that our representations are doomed to be approximations. Finally, it must be recalled that my grounding scheme is proposed in a framework of methodological epiphenomenalism: It only tries to account for performance capacity (discrimination, identification, description), not qualitative experience. So "what it is like to see yellow" is not part of my evidential burden: just what it takes to discriminate, identify and describe colors as those who see yellow do... -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 27 Jun 87 13:22:19 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <917@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > ... blurred the distinction between the > following: (a) the many all-or-none categories that are the real burden > for an explanatory theory of categorization (a penguin, after all, be it > ever so atypical a bird, ... is, after all, indeed a bird, and we know > it, and can say so, with 100% accuracy every time, .... > ... and (b) true "graded" categories such as "big," "intelligent," ... > ...... > "games" are either (i) an all-or-none category, i.e., there is a "right" or > "wrong" of the matter, and we are able to sort accordingly, ... > ... or (ii) "games" > are truly a fuzzy category, in which membership is arbitrary, > uncertain, or a matter of degree. But if the latter, then games are > simply not representative of the garden-variety all-or-none > categorization capacity that we exercise when we categorize most > objects, such as chairs, tables, birds.... Now, much of this discussion is out of my field, but (a) I would like to share in the results, and (b) I understand membership in classes like "bird" and "chair." I learned recently that I can't categorize chairs with 100% accuracy. A chair used to be a thing that supported one person at the seat and the back, and a stool had no back support. Then somebody invented a thing that supported one person at the seat, the knees, but not the back, and I didn't know what it was. As far as my sensory categorization was concerned at the time, its distinctive features were inadequate to classify it. Then somebody told me it was a chair. Its membership in the class "chair" was arbitrary. Now a "chair" in my lexicon is a thing that supports the seat and either the back or the knees. Actually, I think I perceive most chairs by recognizing the object first as a familiar thing like a kitchen chair, a wing chair, etc., and then I name it with the generic name "chair." I think Harnad would recognize this process. The class is defined arbitrarily by inclusion of specific members, not by features common to the class. It's not so much a class of objects, as a class of classes.... If that is so, then "bird" as a categorization of "penguin" is purely symbolic, and hence is arbitrary, and once the arbitrariness is defined out, that categorization is a logical, 100% accurate, deduction. The class "penguin" is closer to the primitives that we infer inductively from sensory input. But the identification of "penguin" in a picture, or in the field, is uncertain because the outlines may be blurred, hidden, etc. So there is no place in the pre-symbolic processing of sensory input where 100% accuracy is essential. (This being so, there is no requirement for invertibility.) M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 28 Jun 87 17:52:03 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: The symbol grounding problem: Against Rosch & Wittgenstein marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel asks: > Why require 100% accuracy in all-or-none categorizing?... I learned > recently that I can't categorize chairs with 100% accuracy. This is a misunderstanding. The "100% accuracy" refers to the all-or-none-ness of the kinds of categories in question. The rival theories in the Roschian tradition have claimed that many categories (including "bird" and "chair") do not have "defining" features. Instead, membership is either fuzzy or a matter of degree (i.e., percent), being based on degree of similarity to a prototype or to prior instances, or on "family resemblances" (as in Wittgenstein on "games"), etc.. I am directly challenging this family of theories as not really providing a model for categorization at all. The "100% accuracy" refers to the fact that, after all, we do succeed in performing all-or-none sorting and labeling, and that membership assignment in these categories is not graded or a matter of degree (although our speed and "typicality ratings" may be). I am not, of course, claiming that noise does not exist and that errors may not occur under certain conditions. Perhaps I should have put it this way: Categorization preformance (with all-or-none categories) is highly reliable (close to 100%) and MEMBERSHIP is 100%. Only speed/ease of categorization and typicality ratings are a matter of degree. The underlying representation must hence account for all-or-none categorization capacity itself first, then worry about its fine-tuning. This is not to deny that even all-or-none categorization may encounter regions of uncertainty. Since ALL category representations in my model are provisional and approximate (relative to the context of confusable alternatives that have been sampled to date), it is always possible that the categorizer will encounter an anomalous instance that he cannot classify according to his current representation. The representation must hence be revised and updated under these conditions, if ~100% accuracy is to be re-attained. This still does not imply that membership is fuzzy or a matter of degree, however, only that the (provisional "defining") features that will successfully sort the members must be revised or extended. The approximation must be tightened. (Perhaps this is what happened to you with your category "chair.") The models for the true graded (non-all-or-none) and fuzzy categories are, respectively, "big" and "beautiful." > The class ["chair," "bird"] is defined arbitrarily by inclusion > of specific members, not by features common to the class. It's not so > much a class of objects, as a class of classes.... If that is so, > then "bird" as a categorization of "penguin" is purely symbolic, and > hence is arbitrary, and once the arbitrariness is defined > out, that categorization is a logical, 100% accurate, deduction. > The class "penguin" is closer to the primitives that we infer > inductively [?] from sensory input... But the identification of > "penguin" in a picture, or in the field, is uncertain because the > outlines may be blurred, hidden, etc. So there is no place in the > pre-symbolic processing of sensory input where 100% accuracy is > essential. (This being so, there is no requirement for invertibility.) First, most categories are not arbitrary. Physical and ecological contraints govern them. (In the case of "chair," this includes the Gibsonian "affordance" of whether they're something that can be sat upon.) One of the constraints may be social convention (as in stipulations of what we call what, and why), but for a categorizer that must learn to sort and label correctly, that's just another constraint to be satisfied. Perhaps what counts as a "game" will turn out to depend largely on social stipulation, but that does not make its constraints on categorization arbitrary: Unless we stipulate that "gameness" is a matter of degree, or that there are uncertain cases that we have no way to classify as "game" or "nongame," this category is still an all-or-none one, governed by the features we stipulate. (And I must repeat: Whether or not we can introspectvely report the features we are actually using is irrelevant. As long as reliable, consensual, all-or-none categorization performance is going on, there must be a set of underlying features governing it -- both with sensory and more abstract categories. The categorization theorist's burden is to infer or guess what those features really are.) Nor is "symbolic" synonymous with arbitrary. In my grounding scheme, for example, the primitive categories are sensory, based on nonsymbolic representations. The primitive symbols are then the names of sensory categories; these can then can go on to enter into combinations in the form of symbolic descriptions. There is a very subtle "entry-point" problem in investigating this bottom-up quasi-hierarchy, however: Is a given input sensory or symbolic? And, somewhat independently, is its categorization mediated by a sensory representation or a symbolic one (or both, since there are complicated interrelations [especially inclusion relations] between them, including redundancies and sometimes even incoherencies)? The Roschian experimental and theoretical line of work I am criticizing does not attempt to sort any of this out, and no wonder, because it is not really modeling categorization performance in the first place, just its fine tuning. As to invertibility: I must again repeat, an iconic representation is only analog in the properties of the sensory projection that it preserves, not those it fails to preserve. Just as our successful all-or-none categorization performance dictates that a reliable feature set must have been selected, so our discrimination performance dictates the minimal resolution capacity and invertibility there must be in our iconic representations. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: Sun 28 Jun 87 15:27:22-PDT From: Ken Laws Subject: Fuzzy Symbolism From: mind!harnad@princeton.edu (Stevan Harnad) Finally, and perhaps most important: In bypassing the problem of categorization capacity itself -- i.e., the problem of how devices manage to categorize as correctly and successfully as they do, given the inputs they have encountered -- in favor of its fine tuning, this line of research has unhelpfully blurred the distinction between the following: (a) the many all-or-none categories that are the real burden for an explanatory theory of categorization (a penguin, after all, be it ever so atypical a bird, and be it ever so time-consuming for us to judge that it is indeed a bird, is, after all, indeed a bird, and we know it, and can say so, with 100% accuracy every time, irrespective of whether we can successfully introspect what features we are using to say so) and (b) true "graded" categories such as "big," "intelligent," etc. Let's face the all-or-none problem before we get fancy... Is a mechanical rubber penguin a penguin? Is a dead or dismembered penguin a penguin? How about a genetically damaged or altered penguin? When does an penguin embryo become a penguin? When does it become a bird? I think your example depends on circularities inherent in our use of natural language. I can't unambiguously define the class of penguins, so how can I be 100% certain that every penguin is a bird? If, on the other hand, we are dealing only in abstractions, and the only "penguin" involved is a idealized living adult penguin bird, then the question is a tautology. We would then be saying that we are 100% certain that our abstraction satisfies its own sufficient conditions -- and even that could change if scientists someday discover incontrovertible evidence that penguins are really fish. In short, every category is a graded one except for those that we postulate to be exact as part of their defining characteristics. After writing the above, I saw the following reply: I am not, of course, claiming that noise does not exist and that errors may not occur under certain conditions. Perhaps I should have put it this way: Categorization preformance (with all-or-none categories) is highly reliable (close to 100%) and MEMBERSHIP is 100%. Only speed/ease of categorization and typicality ratings are a matter of degree. The underlying representation must hence account for all-or-none categorization capacity itself first, then worry about its fine-tuning. This is not to deny that even all-or-none categorization may encounter regions of uncertainty. Since ALL category representations in my model are provisional and approximate (relative to the context of confusable alternatives that have been sampled to date), it is always possible that the categorizer will encounter an anomalous instance that he cannot classify according to his current representation. The representation must hence be revised and updated under these conditions, if ~100% accuracy is to be re-attained. This still does not imply that membership is fuzzy or a matter of degree, however, only that the (provisional "defining") features that will successfully sort the members must be revised or extended. The approximation must be tightened. You are entitled to such an opinion, of course, but I do not accept the position as proven. We do, of course, sort and categorize objects when forced to do so. At the point of observable behavior, then, some kind of noninvertible or symbolic categorization has taken place. Such behavior, however, is distinct from any of the internal representations that produce it. I can carry fuzzy and even conflicting representations until -- and often long after -- the behavior is initiated. Even at the instant of commitment, my representations need be unambiguous only in the implicit sense that one interpretation is momentarily stronger than the other -- if, indeed, the choice is not made at random. It may also be true that I do reduce some representations to a single neural firing or to some other unambiguous event -- e.g., when storing a memory. I find this unlikely as a general model. Coarse coding, graded or frequency encodings, and widespread activation seem better models of what's going on. Symbolic reasoning exists in pure form only on the printed page; our mental manipulation even of abstract symbols is carried out with fuzzy reasoning apparatus. -- Ken Laws ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Mon Jun 29 03:31:18 1987 Date: Mon, 29 Jun 87 03:31:07 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #159 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Mon, 29 Jun 87 03:25 EDT Received: from relay.cs.net by RELAY.CS.NET id ab11152; 29 Jun 87 2:38 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa22302; 29 Jun 87 2:37 EDT Date: Sun 28 Jun 1987 23:05-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #159 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Monday, 29 Jun 1987 Volume 5 : Issue 159 Today's Topics: Future Directions - Drexler and Nanotechnology, History - AI in the 13th Century & Otto Selz, Binding - Computer Composition of Music ---------------------------------------------------------------------- Date: Mon, 22 Jun 87 11:03:03 EDT From: Bruce Nevin Subject: nano-engineering There is a good summary article in _Whole Earth Review_ (Spring 1987), pp. 8-14: A Technology of Tiny Things, Nanotechnics and Civilization, by K. Eric Drexler. The bio in the footnote at the beginning says Drexler got his SB from MIT in interdisciplinary science, followed by a Master's in Aeronautics and Astronautics also at MIT. Recently he founded the MIT Nanotechnology Study Group to develop the science described in the article and book. Some excerpts from the former: Whatever is, is obviously possible. Life is. Therefore that demonstrates the possibility of molecular machines able to build other molecular machines--the essence of both life and a new method called nanotechnology. . . . Whatever obeys natural law is also possible. Science now understands the laws of ordinary matter and energy well enough for most engineering purposes. Nanotechnology will enable us to build new kinds of things. Physical laws let us calculate what some of these things will be able to do. The basic idea of nanotechnology is straightforward. . . . Molecular machines are simply machines made of molecular-scale parts having carefully arranged atoms. . . . Nanotechnology assemblers will be molecular machines that grab reactive molecules and bring them together in a controlled way, building up a complex structure a few atoms at a time. . . . There is no new science in nanotechnology, only new engineering. The possibility of nanotechnology was implicit in the science known over 30 years ago, though no one saw it then. During the 1940s and 1950s, biochemistry revealed more and more of the molecular machinery of the cell. In 1959, physicist Richard Feynman touched on a similar idea in a talk: he spoke of using small machines to build smaller machines ( . . . and so on). He suggested that the smallest machines would be able to "put atoms down where a chemist says" to make a "chemical substance." But Feynman didn't explain how these machines were to work, and said they "will really be useless," because chemists will be able to make whatever they want without them. Decades passed with little followup. [Molecular biology advanced, Drexler's work at MIT indicated in winter of 1976 the possibility of "what we now call assemblers"; he describes several paths for evolution of nanotechnics from present science and technology. --BN] As you can see, the starting point will make little difference. All roads lead to assemblers, and assemblers will let us make almost anything we are clever enough to design. . . . In a world full of competing companies and governments, only global disaster or global domination could block the advance of technology. This seems to be a fundamental principle; if so, it must guide our plans. . . . What can nanotechnology do for us? Almost anything we want, in physical terms. Once we have the software to direct them, replicating assemblers can build almost anything, including more of themselves, without human labor. Because they will handle matter atom by atom, as trees do, they can be as clean as trees, or cleaner. They need not produce smoke or sludge or toxic chemical byproducts. . . . One important application will be the further miniaturization of computers. Detailed study shows that assemblers could build the equivalent of a large, modern computer in about 1/1000 of the volume of a typical human cell. This could be a mechanical computer (they're easier to analyze than electronic computers), but moving parts on this scale can be small and fast enough to make the computer faster than today's electronic machines. . . . Drexler also writes at some length about the enormous potential for danger and disruption of society and biosphere. Our survival may depend on our ability to tell sense from nonsense regarding a complex technology that doesn't exist yet. The nonsense will be abundant, no matter what we do: any field on the borders of science fiction, quantum mechanics, and biology is well positioned to import a lot of prefabricated crap; any field where experiments and experience aren't yet possible is going to have great trouble getting rid of that crap. When someone says "nanotechnology" and begins to expound, beware! . . . a political movement to deal with nanotechnology must be a movement to guide advance, not to stop it. I've already argued that attempts to stop it would be futile; here are some reasons for thinking such efforts would be socially irresponsible. I leave this and much more for the interested reader to follow up in the Spring issue of WER. (This same issue by the way has Shank's `Reality Club' contribution on why math should not be taught in public schools. As you know from his AI work, it cannot be because he dislikes math or is bad at it.) Bruce Nevin bn@cch.bbn.com (This is my own personal communication, and in no way expresses or implies anything about the opinions of my employer, its clients, etc.) ------------------------------ Date: 24 Jun 87 15:39 PDT From: JJD.MDC@OFFICE-1.ARPA Subject: Drexler and Nanotechnology Sorry I missed the NPR report on K. Eric Drexler and _Engines of Creation_. Here's some background: The book _Engines of Creation_ was publixhed by Anchor Press / Doubleday in 1986. The excited foreword is by Marvin Minsky. I know from the copy that I just checked out that Drexler discusses AI in the book, but I am not sure what his vantage point is. At minimum, of course, is the potential of nanotechnology as a way to build much denser hardware. I suspect that Drexler also at least touches on the idea of this enabling a critical mass and consciousness. This will come later. It's a good read. The book was excerpted in the Spring 1987 issue of _Whole Earth Review_. This article is provocative and has some good conceptual illustrations. The cover bio of Drexler identifies him as a "Research Affiliate at the MIT Space Systems Laboratory." He is pictured with his back to the camera, staring at an imposing and eclectic pile of books and a terminal. I first encountered Drexler in the pages of the Summer 1976 issue of _CoEvolution Quarterly_ (ancestor of _Whole Earth Review_). The theme of that issue (later extended as a book) was Gerard O'Neill's concept of space colonization and industrialization. Drexler was a very articulate advocate who was actually doing something about it. He was a graduate student at the time, and had built a six-foot-long track that could electromagnetically launch a bucket of water into a wall at 80 miles per hour. It was a demonstration of the feasibility of a mass driver to be built on the moon and to launch 10 kilogram sacks of material to colony construction sites. He contributed to the book that came from the initial article, demolishing his opponents with gleeful arrogance (apparently since moderated, perhaps by exposure to audiences in the last few years). I suspect that Drexler has a general interest in big fixes in answer to contemporary dilemmas, motivating his fascination with both space industrialization and with nanotechnology. More specifically, his earlier interest in space has probably driven his interest in nanotechnology. Nanotechnology promises to revolutionize both the ways things are made, and their resulting performance characteristics. It makes vast systems of capital-intensive, high-performance technology seems more approachable. I leave further analysis to the next century's graduate theses re: contemporary intellectual history. ------------------------------ Date: 26 Jun 87 18:28:37 GMT From: jbn@glacier.stanford.edu (John B. Nagle) Subject: Re: AI in the 13th Century A thorough discussion of the Ars Magna ("Great Art") of Ramon Lull can be found in Martin Gardner's "Science - Good, Bad and Bogus" (ISBN 0-87975-144-4). The Great Art is basically a system for exhaustively combining terms, using a stack of disks, each containing a set of related terms. For example, one set of Lull's disks contained the following words: 1. God, creature, operation 2. difference, similarity, contrariety 3. beginning, middle, end 4. majority, equality, minority 5. affirmation, negation, doubt In operation, one chooses one term from each set, more or less at random. One can thus explore, Gardner writes, "such topics as the beginning and end of God, differences and similarities of animals, and so on." The Great Art provides no assistance in selecting useful combinations from the many produced, or for doing anything with them once selected. It provides only a means for enumerating the possibilities inherent in some taxonomic scheme. So, while the Great Art may be useful as a prod for creative thinking by humans, it does not provide anything more profound. It does, though, generate the illusion of profundity, which provides much of its appeal. John Nagle ------------------------------ Date: 24 Jun 87 13:23:51 GMT From: edwards@unix.macc.wisc.edu (mark edwards) Subject: AI in the 13th Century A number of people have asked about the reference to AI in the 13th Century. Well I finally dug up the ole notebook and picked it out. Unfortunately all I have is a name. The name is Ramon Lull Since the book was in latin, very old and so forth I guess I thought I'd never check it out. Apparently Ramon was a popular person in the sciences, black magic and those sort of things. His name appears with other terms like shamans in my notebook. I hope that helps. mark -- edwards@unix.macc.wisc.edu {allegra, ihnp4, seismo}!uwvax!uwmacc!edwards UW-Madison, 1210 West Dayton St., Madison WI 53706 ------------------------------ Date: 24 Jun 87 17:03:34 GMT From: duke!mps@mcnc.org (Michael P. Smith) Subject: Re: AI in the 13th Century In article <1654@uwmacc.UUCP> edwards@uwmacc.UUCP (mark edwards) writes: > > A number of people have asked about the reference to AI in the > 13th Century. Well I finally dug up the ole notebook and picked > it out. Unfortunately all I have is a name. The name is > > Ramon Lull > > > Since the book was in latin, very old and so forth I guess I thought > I'd never check it out. Apparently Ramon was a popular person in the > sciences, black magic and those sort of things. His name appears > with other terms like shamans in my notebook. > I'm no Lull expert, but here's part of an entry from W.L. Reese's DICTIONARY OF PHILOSOPHY AND RELIGION (Humanities, 1980), p. 319: \begin{quotation} Lull, Raymond. 1236-1315. Philosopher and missionary. Born in Palma, Majorca. Taught several years at Paris. His goal was to state the truths Christianity so succinctly that the infidels could not possibly deny them. To this end he wrote the *Ars Magna*, a mechanical method of exhaustively stating the possible relations of a topic. The method requires three concentric circles divided into compartments. One circle is divided into nine relevant subjects; a second circle is divided into nine relevant predicates; the third circle is divided into nine questions: whether? what? whence? why? how large? of what kind? when? where? how? One circle is fixed; the others rotate, providing a complete series of questions, and of statements in relation to them. \end{quotation} Lull is usually dismissed as a crackpot by historians, but had influence on the likes of Descartes and Leibniz centuries later. I believe that much of Lull's work is available in English translation. No doubt some interesting comparisons can be drawn between Lull's program and, say, conceptual dependency theory. But as to Mark's claim that Lull used the term 'artificial intelligence', I suspect that such usage occurs only in the mind of the translator. ---------------------------------------------------------------------------- Michael P. Smith "The world of the happy man is a different ARPA: mps@duke.cs.duke.edu one from that of the unhappy man." Wittgenstein ------------------------------ Date: 24 Jun 87 19:05:25 GMT From: duke!jds@mcnc.org (Joseph D. Sloan) Subject: Re: AI in the 13th Century Martin Gardner devotes a chapter to Ramon Lull in LOGIC, MACHINES AND DIAGRAMS, 2e, 1982, University of Chicago Press. Joe Sloan jds@duke ------------------------------ Date: 25 Jun 87 14:03:37 GMT From: nosc!humu!uhccux!stampe@sdcsvax.ucsd.edu (David Stampe) Subject: Re: AI in the 13th Century The nine questions of Ramon Lull's Ars Magna (whether? what? whence? why? how large? of what kind? when? where? how?) seem to be what were called the "modes of being" in the grammatical theories of the "Modistae" during the middle ages. They were based ultimately on Aristotle's Categories, which have been claimed during this century (by Ryle?) to have been based on the Greek interrogative pronouns. Regarding the similarities to conceptual dependency theory, it's interesting that in *syntactic* dependency theory, in a phrase, it is only the dependent member (adjunct, modifier, operator) that can be interrogated vis a vis the independent (head, operand) member, not vice versa. Examples, with (Head (Adjunct)), and * for the bad cases: (Verb (Object)) Q: Who does he like? A: Mary. *Q: What he Mary? A: Likes her. ((Adj) Noun) Q: Which hat did she wear? A: The straw hat. *Q: What straw did she wear? A: The hat. ((Adv) Adj) Q: How hot was it? A: Too hot. *Q: Too what was it? A: Hot. Etc. Typically the head is implied by the adjunct (e.g. to like Mary is to like [someone], a straw hat is a hat, too hot is hot). That is, adjuncts are rather like predicates. That is, they correspond to the modes of being, the ways things can be. There's not much new under the sun. David Stampe, Linguistics, Univ. of Hawaii uhccux!stampe@nosc.mil ------------------------------ Date: Sun, 21 Jun 87 19:24:42 +0300 From: NYSTERN%WEIZMANN.BITNET@wiscvm.wisc.edu Subject: Re: Re: Taking AI models and applying them to biology... I have two comments to say; a) As far as I understand from the article, Otto Selz has DIED in 1943 at Auschwits (If one takes into account what Auschwits was that period it seems quite logical) thus he couldn't publish his work in 1943 ... If one remembers what were the types of people who died in Auschwits (I.E Jews) and if one takes into account that they expelled from The Universities and Research Institutions from about 1933 (Hitlers election as Germany's prime minister) then the only logical concultion is that He didn't publish his theory after 1933 (since he was banned) 1943-1933=10 years (woow , I made it ...) which means he published his theory before Turing or Shannon published theirs ... WWII was probably the main reason for the lack of knowledge about his work. (Remember that the war was ended 2 years later and the world had enough on his mind than to remember Selz's theory ...) BTW The commentation above isn't based on facts since I know very little about his life and death (I may be wrong and will found out that he died As a top Nazy SS officer due to cancer ... but that possiblity seems redundant to me) b) As far as I know AI is based on Mathematics ans Biology. Both of those Sciences and many of the disciplines adopted by AI scientists were formed a long time ago even without being influenced by AI/computers (in matter of fact up until now the fields of Biology and Computers wasn't combined together when matters of theory comes only as a tool (calculation programs and DNA decoding algorythms) Well to sum up my point I feel that the computer-science field will benefit more from the work of hopfield then any theoretical axiome ... The scince world has become too specific while I believe that combining all forces together instead of working in paralel on the same topic would be more fruitful for the science world and for the world (With one objection that one field should not impose his theory on the other let, 1000 flowers grow together, but TOGETHER). There is a Master grad. in weizmann who has done his thesis in the vision field in the Department of Applied Math, His 'problem' is that his thesis relates to many fields (physics,neuro biology) and not only to Applied Math moreover He has proved emphirically and cited Famous researchers in this field that Math is redundant in this specific field. Ofcourse no one like his thesis in the Math Department ... As far as I know He will have his Master (got above 80 in the oral test) BUT how much will he get about hsi work ? no one knows. His work is great but it doesn't fit into the cateogries of our formal science. There's no (yet) a field named Applied neuro-math or Applied psycho-physics or even Applied neuro-physics. I've brought this story up to show 1) The situation in the science world nowadays 2) to emphesize the trends of science as I see them 3) to back up the notion that Applied Math and AI would benifit alot by examening works of other science fields. I believe that that's all. ------------------------------ Date: 21 Jun 87 19:42:03 GMT From: sunybcs!rapaport@ames.arpa (William J. Rapaport) Subject: Re: Computer composition of music In article <2198@mmintl.UUCP> johnt@mmintl.UUCP (John Tangney) writes: >Some of the researchers I read about (like Max >Mathews, Lejaren Hiller, Iannis Xenakis, Stephen Smoliar to name >a few off the top of my head) must still be out there. Lejaren Hiller is Prof. of Music at SUNY Buffalo and an adjunct prof. in our CS dept. His email address is muslah@buffalo.csnet ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 30 08:23:41 1987 Date: Tue, 30 Jun 87 08:23:33 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #160 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 30 Jun 87 08:10 EDT Received: from relay.cs.net by RELAY.CS.NET id ac17446; 30 Jun 87 2:06 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa02809; 30 Jun 87 2:05 EDT Date: Mon 29 Jun 1987 22:17-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #160 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 30 Jun 1987 Volume 5 : Issue 160 Today's Topics: Queries - Plausible Reasoning & Natural Language - Predicate Calculus - Theorem Proving & Automatic Programming Bibliographies & Frame Matching and Chaining, Psychology - Why Did The $6,000,000 Man Run So Slowly? ---------------------------------------------------------------------- Date: Tue, 23 Jun 87 20:57:05 SST From: Jenny Subject: so what about plausible reasoning ? As I read articles on plausible reasoning in expert systems, I come to the conclusion that experts themselves do not exactly work with numbers as they solve problems. And many of them are not willing to commit themselves into specifying a figure to signify their belief in a rule. The deductive process that occurs in their brain can never be replicated by any known plausible reasoning models. The expert system technology is already a weak one per se, why introduce further complexity and more bottleneck in the acquisition of knowledge, knowing fully well that the numbers are probably inconsistent ? If one obtains two conclusions with numbers indicating some significance, say 75 % and 80 %, can one say that the conclusion with 80% significance is the correct conclusion and ignore the other one ? These numbers do not seem to mean much since they are just beliefs or probabilties. Lim Eng-Lian National University of Singapore -- this opinion is my own and is not influenced by the color of my office ------------------------------ Date: Wed, 24 Jun 87 16:37:04 SET From: "Adlassnig, Peter" Subject: natural language - predicate calculus - theorem proving concerning my ph.d. thesis i would like to know who has dealt already with the following themes: 1) translation of indefinite pronomina into predicat calculus. parsing only simple english sentences (subj pred obj), with reference to the distribution and interpretation of wh-words and quantifiers. ( a lexicon should be minimized to the syntax and not include semantik ambiguities-rules.) 2) representation of quantifiers in frames. 3) automated theorem prover algorith, which is easy to implement for first-order predicat-logic. are logic grammars the right field for 1)? the aim of the whole system is to implement an expertsystem in logo, to demonstrate in schools, that computers can "think". i would be thankful for any help| which literature would you advice to read? ruth gruenberger Please send response to adlassni%awiimc11.bitnet Thank you Peter Adlassnig ------------------------------ Date: 29 Jun 87 19:36:07 GMT From: pratt@vanhalen.rutgers.edu (Lorien Y. Pratt) Subject: Request for automatic programming bibliographies Has anyone recently put together a bibliography of work in automatic program generation? I'd appreciate any pointers that you can give me. --Lorien Pratt ------------------------------ Date: 23 Jun 87 21:13:20 GMT From: ihnp4!drutx!mcp@ucbvax.Berkeley.EDU (Mike Paugh) Subject: Need info on frame matching and chaining I am looking for good reference material on building expert systems using frames in Lisp. The environment will be GCLISP. What I need is a good basic understanding of how to chain through the frames and do the pattern matching. I am new to this, so any good pointers will be appreciated. Mike Paugh AT&T IS Labs Denver ihnp4!drutx!mcp ------------------------------ Date: 29 Jun 87 12:34:12 GMT From: ihnp4!homxb!mtuxo!mtune!akgua!cpsc53!dwb@ucbvax.Berkeley.EDU (Summer Hire) Subject: Re: Need info on frame matching and chaining > > I am looking for good reference material on building expert > systems using frames in Lisp. The environment will be GCLISP. > What I need is a good basic understanding of how to chain > through the frames and do the pattern matching. > > I am new to this, so any good pointers will be appreciated. > > > Mike Paugh > AT&T IS Labs Denver > ihnp4!drutx!mcp Hi, I am a summer hire from AT&T, and just finished a two term course on conceptual dependencies (Frame-style inference netting) and pattern matching. We conducted six member projects on the building of MARGIE. This Margie took an english story and converted it into "Frames" where then it was pattern matched against set scripts (senerios or events). From this matching the system could then construct an infered sequence of actions form the given. It took about six months to develop. The book we used as an outline, which gave us a firm grasp on the basics of the entire system, was Roger Shanks book named "INSIDE COMPUTER UNDERSTANDING." It gave us a great lead. We did vary to a certain extent in actual development, but the basics are still there. I would be glad to aid you in any way for those further developments, because they provided a more natural way of demonstrating working cognitive structures. (In theory of course.) I must say that the book is a must to get you going. Another great lead in this search would be to contact George Stockman, Prof. at Michigan State University. He was the developer of our course project and is a biggie on frame representation of knowledge and expert systems. (He taught me every thing I know). Please contact me if you need any assistance at all. Dave Bigelow (summer hire and damn well worth it!) ------------------------------ Date: 20 Jun 87 15:44:33 GMT From: mit-amt!mob@mit-amt.media.mit.edu (Mario O. Bourgoin) Reply-to: mit-amt!mob@media-lab.media.mit.edu (Mario O. Bourgoin) Subject: Re: Why Did The $6,000,000 Man Run So Slowly? Because it made the special effects scenes last longer. ------------------------------ Date: Mon 22 Jun 87 10:12:25-CDT From: Art Flatau Subject: Re: why did the $6,000,000 man run so slowly? I think people have missed the obvious reason that the $6 Meg man ran so slowly. To stretch the plots out to fill an hour time slot. Art ------------------------------ Date: Mon, 22 Jun 87 09:47:32 PDT From: lambert%cod@nosc.mil Subject: Why did $6M man run so slowly? Re: Why did $6M man run so slowly? Why would a producer use slow motion to depict very fast movement? I suggest the following reasons be added to the list: 1. ACHIEVE THE TECHNICAL EFFECT. The slow motion points out to the viewer the fact that the flow of time is different. The context around the slow-motion scene makes the magnitude and direction of this change obvious. This is all the viewer really needs to sense the effect that the $6M man is moving much faster than normal. 2. TAKE ADVANTAGE OF THE VIEWER'S IMAGINATION. The slow motion gives the viewer's mind time to realize that fast motion is being represented, and to appreciate the non-triviality of it (unlike a realtime presentation which would tend to make it seem easy). It allows the viewer's imagination to be creative, to draw on previous experience, and to construct the concepts and images necessary to represent something so complex and magnificently-engineered happening so fast. This increases the impact on the viewer by enhancing appreciation of the $6M man's feats. Indeed, it can give the viewer an experience far beyond what the producer can actually achieve on the screen. 3. TAKE ADVANTAGE OF THE VIEWER'S INTEREST IN LEARNING ABOUT HIMSELF. The viewer is treated to a slow-motion presentation of human qualities difficult or impossible to observe at normal or faster speeds. This allows him to learn new things about the actor, himself, and other humans. 4. ACHIEVE ARTISTIC EFFECT. The producer also achieves beautiful artistic effect by allowing viewing of the visible signs of forces and motion, observation of facial expressions, and contemplation of the beauty of the $6M man's athletic qualities such as speed, power, grace, and coordination. lambert@cod.nosc.mil (David R. Lambert) ------------------------------ Date: Mon, 22 Jun 87 16:38:32 PDT From: "William J. Fulco" Subject: Slow-motion / $6E6 man amsler@flash.bellcore.com: > .... > I suspect what is happening is that this is analogous to the focusing > of attention on the events which happened in a real moving image > memory. That is, if one attempts to reconstruct an event that > happened very quickly in real time after the fact, one will > artificially create something like slow motion. This "slo-motion" effect of perception also appears to work in real-time. A good everyday example of this is (for people that play sports) a pass or "drive" in basketball, a volly in tennis or hitting a baseball. Professional baseball players talk about learning to see the ball they are trying to hit. They say that they actuall see the ball - an object the size of an orange, traveling at 90+ mph from 66 feet away. I used to think that this wasn't really what was happening, but I have been involved in basketball games where, for less than 1 second, (real-time) I have had an open lane to the basket, or an oportunity to make a pass. The perceived time was far slower, on the order of several seconds. During these perceived seconds, I had time to "think" about my options - actually make verbal & image (mind's eye) judgments about what to do or not to do, commit and make or skip the play. One case of this that really stands out: playing basketball several weeks ago I was left wide open for drive to the basket. I remember that I couldn't beleive I was left this wide open and I started to think "what's the catch". I then remember thinking that "I don't have time to be thinking about thinking about what I should be doing - I should just go", and with this I drove down the key (-: missed the shot :-). The point is, I had time to "argue" with myself, "verbally", in my mind before I took action, but in real-time no more that a second passed. The first time you notice this effect it is truly erie. (bill) [Yup. It happened to me once, in 1962, as I was jumping out of a swing into a sandlot. I had done this (at full speed from maximum height) hundreds of times, and did so again afterwards, but only this once did time slow to about 1/4 speed. I wonder if a similar effect might be a part of the "born again" religious conversion that is sometimes hits people during routine activities. -- KIL] ------------------------------ Date: 27 Jun 87 13:41:23 GMT From: winfree!uucp@seismo.CSS.GOV (Unix Chit-Chat at winfree.n3eua.cos.ampr.n3eua.cos.ampr) Subject: Submission for comp-ai-digest Path: winfree!hp-lsd!hpldola!ben From: ben@hpldola.HP.COM (Benjamin Ellsworth) Newsgroups: comp.ai.digest Subject: Re: Why Did The $6,000,000 Man Run So Slowly? Message-ID: <13330001@hpldola.HP.COM> Date: 26 Jun 87 20:42:14 GMT References: <870615144826.2.NICHAEL@BUBBAROMDOS.PALLADIAN.COM> Organization: HP Logic Design Oper. -ColoSpgs Lines: 15 >From my film classes at school, I had gathered that the reason that the action sequences in Kung Fu were slowed down for emphasis. When you slow a scene down, whatever the content, you emphasize the action of that scene. This is especially effective for violent action. Any good anti-hunting film will slow down any shots of an actual Bambi kill. The effect of slowing is to force the viewer to perceive the action in more detail (and hence with greater emphasis) than he/she could view it at normal speed. Speeding up a scene has the opposite effect. Benjamin Ellsworth hplabs!hpldola!ben *** This posting is about the use of temporal distortion in film making, not a statement regarding the morality of hunting. ------------------------------ Date: 24 Jun 87 21:34:36 GMT From: ihnp4!chinet!nucsrl!coray@ucbvax.Berkeley.EDU (Elizabeth) Subject: Re: Why did the six-million dollar man run so slowly? / nucsrl:comp.ai / tim@linc.cis.upenn.edu (Tim Finin) / 11:47 pm Jun 11, 1987 / Why did the six million dollar man run so slowly? The guy moves slowly in the same way that a car accident happens "slowly". Slow motion simulates the increase in attention to detail and reaction times which go with an increase in adrenaline. This makes slow motion, oddly enough, exciting. The thing with the cougar is right on because the pedator in the hunt is just the sort of thing for which adrenaline evolved. M. E. Corey ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 30 08:23:57 1987 Date: Tue, 30 Jun 87 08:23:47 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #161 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 30 Jun 87 08:13 EDT Received: from relay.cs.net by RELAY.CS.NET id ad17446; 30 Jun 87 2:07 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa02814; 30 Jun 87 2:08 EDT Date: Mon 29 Jun 1987 22:26-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #161 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 30 Jun 1987 Volume 5 : Issue 161 Today's Topics: Query - Mega-Monitor, Robotics - Vectrobot Recommendation, AI Tools - Object-Oriented Languages & CLP(R) Announcement ---------------------------------------------------------------------- Date: 26 Jun 87 15:50:15 GMT From: stride!tahoe!unsvax!jimi!asci!brian@gr.utah.edu (Brian Douglass) Subject: Mega-Monitor I've been asked by a friend to research information about a super-size monitor. Essentially, what I am looking for is a color monitor that is 10 feet by 10 feet with a resolution of say 13,000 by 13,000--don't ask me what for because I don't know what for, I'm just the gumshoe--and any necessary equipment to drive it. (can you imagine what kind of equipment is necessary to drive 169 million pixels!) Basically, my friend needs to generate some very large images with extremely fine details. Color is preferable, but not absolute. Money is not really a concern at this moment, so lets hear anything you got. Also, if it means building it custom, that is what my friend wants to know. So IBM, RCA, Tektronik, HP, etc, if you're listening and have any experience in a monitor this large I would like to hear from you, as well as if you have any "off-the-top-of-you-head" price estimates send those along. I fully expect to hear in the millions, but that's okay. Right now, my friend needs to know if anybody has done this, or if anybody can do this. I know that there are also some analog systems out and about of this type of magnitude. Although not preferable, I would like to hear about them as well. Please E-mail only your responses to me as I am gone periodically on business and we keep only a days worth of news on our system for a myriad of reasons. However, I will summarize periodically to comp.graphics the responses I do receive, as I am sure there are others who are as fascinated as I with the leviathan proportions of this Mega-Monitor. Brian Douglass Applied Systems Consultants, Inc. (ASCI) P.O. Box 13301 Las Vegas, NV 89103 Office: (702) 733-6761 Home: (702) 871-8182 brian@asci.uucp UUCP: {akgua,ihnp4,mirror,psivax,sdcrdcf}!otto!jimi!asci!brian ------------------------------ Date: 22 Jun 87 19:49:48 GMT From: linus!alliant!sullivan@husc6.harvard.edu (Mike Sullivan) Subject: Re: Search and Employ (Mobile robot) If your work involves mobile robot navigation, or other research in robotics requiring sturdy, reliable hardware, I recommend a small three wheel drive, three wheel steer (synchronous drive) chassis called a "Vectrobot". It is manufactured by a company in New Hampshire called Real World Interface. For info, or references call Grinnell or Curt at (603) 654-6334 #include ______ / \ \ Michael J Sullivan / \____\ Alliant decvax!linus!alliant!sullivan / / \ ComputerSystemsCorporation /____/_______\ ------------------------------ Date: 21 Jun 87 11:42:01 GMT From: munnari!koel.rmit.oz!rcopm@seismo.CSS.GOV (Paul Menon) Subject: Re: Smalltalk-80 for Sun 3 ... (LONG) In article <8706180728.AA10707@ucbvax.Berkeley.EDU>, lcc.bill@CS.UCLA.EDU ("William J. Fulco") writes: > I saw a really nice system, (I mean REALLY nice - with good color support) > from Xerox PARC marketing spinoff at the 1986 AAAI show. It was running > on a Sun 3/260 and it really sizzles..... I can believe that, I was introduced to the 3/260 just recently. It would make anything sprout wings. The reason for my addition has a little to do with suns, Smalltalk, and technology in general, so please bear with me. This is long. I have just completed some sizable programs (well, to me they were), and am in the "recovery stage", ie sizing up what I have done... was it all worth while etc, etc. I have reached a few (frustrating) conclusions/opinions... * If I leave these programs for a while, and then come back to change them in the name of maintenance or further enhancement, I am not too much better off than someone who has never seen the package before. I don't mean that in the positive sense, nor would I have forgotten the techniques I used; it is the dependence of data being spread all over the program. It was written in Pascal. How many of you decide to change a data structure halfway through a program, not because of bad planning in the first place, but because a "new" and more efficient technique requires extra "bits" embedded into a data structure. Does "grep 'structtype' *.h *.p" ring a bell? Not even then am I too sure if everything is covered, especially if it belongs to an overall complex data structure with crosslinks. No amount of documentation or cross-references will relieve the manual task ahead. Good programming style can minimize this only to a certain extent. C, Algol, Modula 2, perhaps even Ada suffers from this. * If I want to re-use techniques in another program, major surgery is required. Some call this hacking. It's only ok if the same types are being used by the new program. There is static binding available from Ada, if you wish to learn such a complex language. But none of the "standard languages" allow complete type independence. Lisp and Prolog programs will suffer the same scalpel treatment as the others. If you haven't already guessed where I am heading, object-oriented programming languages will (in my opinion) relieve me of these woes. Ok says I, which one do I use? There is the Grand-Daddy, Smalltalk-80; the pure one. Then there are the nouveau hybrids C++, Objective-C and MacApp. Others come in different Flavors, Loops or feathered Flamingos and Owls. Lisp and Prolog do not satisfy my requirements because I cannot easily "build" on previous applications experience. I don't hold the generally dismal performance of Smalltalk aginst it. Hardware is zooming ahead as witnessed on suns, and soon on the Mac II (I hope). My questions to all who have not gone to sleep are... Will Smalltalk mature from being the toy that it was? ie a full 32 bit machine with > 32000 objects etc.. Methinks this is the ideal language to be using no matter how big or small the program. The objection to being such an "open" system can be countered by their "change management tools", if I may be permitted to steal the phrase. Of the hybrids, Objective-C appears my favourite. Although I have never used it, I delight in the similar Smalltalk syntax. It will be a good stand-in until Smalltalk meets its hardware match. Could any user out there please comment on Objective-C, including it's ease of use, availability, portability (ie, which o/s's can it run on), and price? My preference to Objective-C rather than C++ is that I "feel uncomfortable" in the way the latter has been implemented. The extended syntax does not "stand out", it either melds into the other hieroglphs, so I cannot pick the wood from the trees or it further confuses my understanding of C. I wish I had a video of me reading a C program.. I must have this perpetual frown. Is it common? I would love to hear from C++ users, especially those who have used C++ and Objective-C. Note that my primary preference to Objective-C is its syntactical similarity to Smalltalk. Why not MacApp as an interim? Why not indeed! It is another example of brilliance on the part of Apple, and once I get over the confusion of records and messages/methods, all should be swell. One hitch though. Apple had deemed it necessary to inflict a licencing fee on anyone producing/marketing software that uses MacApp, as well as restricting all such programs to the Macintosh. I don't know whether this still holds. I have noted MacApp being used on a 4.2 bsd system (refer to OOPSLA '86 procs pp 186 - 201). pity. My main hope is Smalltalk. It is a pity that the ones that can really benefit from such a system are usually the last to see it. Kids. It is the big kids; ie those who have been ingrained or fed up with procedural languages that get to use it. Does this make us shortsighted? Or perhaps fatally dependent on the past? This isn't a plug for trendy software. This is frustration with writing applications from scratch that use (nearly) the same techniques time and time again. I use the hardware of tomorrow, but give it the brains of yesterday. I am supposed to build on experience; all I do is re-invent the wheel. If you have read the book .. "Object Oriented Programming: An Evolutionary Approach" Brad J. Cox. Then a major part of my article echoes its theme. I could not have read it at a more pertinent time. Thankyou, Paul Menon. Dept of Communication & Electronic Engineering, Royal Melbourne Institute of Technology, 124 Latrobe St, Melbourne, 3000, Australia ACSnet: rcopm@koel UUCP: ...!seismo!munnari!koel.rmit.oz!rcopm CSNET: rcopm@koel.rmit.oz ARPA: rcopm%koel.rmit.oz@seismo BITNET: rcopm%koel.rmit.oz@CSNET-RELAY PHONE: +61 3 660 2619. ------------------------------ Date: Fri, 26 Jun 87 15:18:25 est From: munnari!moncsbruce.oz!clp@seismo.CSS.GOV (The CLP(R) Personae) Subject: CLP(R) Distribution Announcement (Can you please add the following announcement to the digest) DISTRIBUTION NOTICE ___________________ We are pleased to announce the availability of our interpreter for CLP(R), the new Constraint Logic Programming language. This is being distributed in source code written in C and it is compatible with most machines running UNIX, eg. Vaxen, Pyramids and Suns. This is not intended to be a commercial announcement and is targeted at educational or research usage. The distribution includes: 1. CLP(R) interpreter (source code); 2. Example CLP(R) programs; 3. Installation Manual and Programmer's Manual (hard copies). Further information can be found in the following papers: 1. J. Jaffar and J-L. Lassez, "Constraint Logic Programming", Proc. 14th ACM-POPL, Munich, January 1987. 2. J. Jaffar and S. Michaylov, "Methodology and Implementation of a CLP System", Proc. 4th ICLP, Melbourne, May 1987. 3. N.C. Heintze, S. Michaylov and P.J. Stuckey, "CLP(R) and Some Electrical Engineering Problems", Proc. 4th ICLP, Melbourne, May 1987. 4. C. Lassez, K. McAloon and R. Yap, "Constraint Logic Programming and Option Trading", IEEE Expert, Fall Issue 1987, to appear. If you would like a Site licence for educational or research purposes, please send a request for more information to either, (a) Electronic Mail address: ACSNET: clp@moncsbruce.oz ARPANET,CSNET: clp@moncsbruce.oz.au UUCP: seismo!munnari!moncsbruce.oz!clp (b) Paper Mail address: CLP(R) Distribution Department of Computer Science Monash University Clayton Victoria 3168 Australia In order to cover distribution and media costs, a license fee of $150 will apply. ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Tue Jun 30 08:24:17 1987 Date: Tue, 30 Jun 87 08:24:00 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #162 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Tue, 30 Jun 87 08:14 EDT Received: from relay.cs.net by RELAY.CS.NET id ac17505; 30 Jun 87 2:19 EDT Received: from stripe.sri.com by RELAY.CS.NET id ab02860; 30 Jun 87 2:18 EDT Date: Mon 29 Jun 1987 22:32-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #162 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Tuesday, 30 Jun 1987 Volume 5 : Issue 162 Today's Topics: AI Tools - Kyoto Common Lisp ---------------------------------------------------------------------- Date: Mon 22 Jun 87 20:24:22-CDT From: CL.BOYER@R20.UTEXAS.EDU Subject: Kyoto Common Lisp Kyoto Common Lisp (KCL) is a complete implementation of Common Lisp written by T. Yuasa and M. Hagiya working under Professor R. Nakajima at the Research Institute for Mathematical Sciences, Kyoto University. It runs on many different machines and is highly portable. It executes very efficiently and it is superbly documented. KCL is being made available at no fee through the implementors' generosity. The complete sources are included. One channel of distribution is via ftp on the Arpanet/Internet. LICENSE REQUIRED! IMPORTANT: Although there is no fee, KCL is not in the public domain. You are authorized to obtain it only after signing and mailing in a license agreement. Before you ftp KCL files you MUST fill out and send in the license agreement included in this message. Otherwise, you are not permitted to make copies of KCL. COPYING KCL VIA INTERNET KCL may be obtained from Internet source rascal.ics.utexas.edu [128.83.144.1], a Sun-3 at the University of Texas at Austin. To obtain KCL, login as "ftp" with password "guest". There are three tar files: /pub/kcl.tar, 4.0 megabytes /pub/kcl.tar.C, produced by compact from kcl.tar, 2.8 megabytes /pub/kcl.tar.Z, produced by compress from kcl.tar, 1.2 megabytes Any of the three files is sufficient to generate KCL. Please ftp the compressed file if possible. Please use ftp at an odd hour if possible to reduce traffic on a sometimes heavily loaded network. Be sure to use binary mode with ftp. A current version of this message may be found as the file /pub/kcl.broadcast. MACHINES ON WHICH KCL RUNS KCL runs on many machines. With the sources provided in the ftp file, KCL may be executed on the following machines (and operating systems). VAX/UNIX (4.2BSD) SUN2 (OS2, 3) SUN3 (OS3) SONY'S NEWS (4.2BSD) ATT3B2 (System V) Fujitu S3000 (System V) Sumitomo's E15 (Uniplus System V) Data General MV (DGUX) Instructions for making the system are in the file doc/porting in the ftp tar file. KCL LICENSE FORM To obtain the right to copy KCL, sign this license form and send it and a copy to the Kyoto address at the end of the form. ONCE YOU HAVE MAILED THE SIGNED LICENSE FORM, YOU MAY COPY KCL. YOU DO NOT HAVE TO WAIT FOR RECEIPT OF THE SIGNED FORM. --------------------------- cut here ---------------------------- LICENSE AGREEMENT FOR KYOTO COMMON LISP The Special Interest Group in LISP (Taiichi Yuasa and Masami Hagiya) at the Research Institute for Mathematical Sciences, Kyoto University (hereinafter referred to as SIGLISP) grants to USER NAME: _________________________________________ USER ADDRESS: ______________________________________ ______________________________________ (hereinafter referred to as USER), a non-transferable and non-exclusive license to copy and use Kyoto Common LISP (hereinafter referred to as KCL) under the following terms and conditions and for the period of time identified in Paragraph 6. 1. This license agreement grants to the USER the right to use KCL within their own home or organization. The USER may make copies of KCL for use within their own home or organization, but may not further distribute KCL except as provided in paragraph 2. 2. SIGLISP intends that KCL be widely distributed and used, but in a manner which preserves the quality and integrity of KCL. The USER may send a copy of KCL to another home or organization only after either receiving permission from SIGLISP or after seeing written evidence that the other home or organization has signed this agreement and sent a hard copy of it to SIGLISP. If the USER has made modifications to KCL and wants to distribute that modified copy, the USER will first obtain permission from SIGLISP by written or electronic communication. Any USER which has received such a modified copy can pass it on as received, but must receive further permission for further modifications. All modifications to copies of KCL passed on to other homes or organizations shall be clearly and conspicuously indicated in all such copies. Under no other circumstances than provided in this paragraph shall a modified copy of KCL be represented as KCL. 3. The USER will ensure that all their copies of KCL, whether modified or not, carry as the first information item the following copyright notice: (c) Copyright Taiichi Yuasa and Masami Hagiya, 1984. All rights reserved. Copying of this file is authorized to users who have executed the true and proper "License Agreement for Kyoto Common LISP" with SIGLISP. 4. Title to and ownership of KCL and its copies shall at all times remain with SIGLISP and those admitted by SIGLISP as contributors to the development of KCL. The USER will return to SIGLISP for further distribution modifications to KCL, modifications being understood to mean changes which increase the speed, reliability and existing functionality of the software delivered to the USER. The USER may make for their own ownership and use enhancements to KCL which add new functionality and applications which employ KCL. Such modules may be returned to SIGLISP at the option of the USER. 5. KCL IS LICENSED WITH NO WARRANTY OF ANY KIND. SIGLISP WILL NOT BE RESPONSIBLE FOR THE CORRECTION OF ANY BUGS OR OTHER DEFICIENCIES. IN NO EVENT SHALL SIGLISP BE LIABLE FOR ANY DAMAGES OF ANY KIND, INCLUDING SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF KCL. 6. This license for KCL shall be effective from the date hereof and shall remain in force until the USER discontinues use of KCL. In the event the USER neglects or fails to perform or observe any obligations under this Agreement, this Agreement and the License granted hereunder shall be immediately terminated and the USER shall certify to SIGLISP in writing that all copies of KCL in whatever form in its possession or under its control have been destroyed. 7. Requests. KCL is provided by SIGLISP in a spirit of friendship and cooperation. SIGLISP asks that people enjoying the use of KCL cooperate in return to help further develop and distribute KCL. Specifically, SIGLISP would like to know which machines KCL gets used on. A brief notice form is appended to this agreement which the user is requested to send by email or otherwise. Please send in further notifications at reasonable intervals if you increase the number and type of machines on which KCL is loaded. You may send these notices to another USER which is cooperating with SIGLISP for this purpose. USER DATE: _________________________________________ BY: ___________________________________________ TITLE: ________________________________________ ADDRESS: ______________________________________ ______________________________________ SIGLISP DATE: _________________________________________ BY: ___________________________________________ Taiichi Yuasa Masami Hagiya Special Interest Group in LISP Research Institute for Mathematical Sciences Kyoto University Kyoto, 606, JAPAN Telex: 05422020 RIMS J JUNET: siglisp@kurims.kurims.kyoto-u.junet CSNET: siglisp%kurims.kurims.kyoto-u.junet@utokyo-relay.csnet USER has loaded KCL on the following machines since (date): Model Number Production Name Number of Machines END OF LICENSE FORM --------------------------- cut here ------------------------ DOCUMENTATION The principal documentation for KCL is, of course, the book "Common Lisp The Language" by Guy L. Steele, Jr. with contributions by Scott E. Fahlman, Richard P. Gabriel, David A. Moon, and Daniel L. Weinreb, Digital Press, 1984. Implementation-specific details of KCL (debugging, garbage collection, data structure format, declarations, operating system interface, installation) may be found in the 131 page "Kyoto Common Lisp Report" by Taiichi Yuasa and Masami Hagiya, the authors of KCL. This report is available from: Teikoku Insatsu Inc. Shochiku-cho, Ryogae-cho-dori Takeya-machi Sagaru, Naka-gyo-ku, Kyoto, 604, Japan tel: 075-231-4757 for 5,000 yen plus postage. The KCL Report is produced by the text-formatter KROFF (Kyoto ROFF), which is used locally within Kyoto University. Currently KROFF works only on printers available in Japan. It is possible that an American distributorship of this report will be arranged. The source of the report, with KROFF commands, is found in the file doc/report on the ftp tar file. It is possible to read this source, though it is as hard on the eyes as TeX or Scribe source. A translation of this source into TeX is underway and will be available as part of the distribution tape. Future information about the availability of the KCL Report will be available in updated versions of this message, in the file /pub/kcl.broadcast. A document describing how to port KCL to other systems is available at no charge from the authors of KCL. Each of the KCL primitives is thoroughly described by the "describe" function, which is based on 340K bytes of documentation. SUPPORT KCL is one of the most bug-free large software systems that we have ever used. However, when bugs are found, they may be reported to the implementors: hagiya%kurims.kurims.kyoto-u.junet%utokyo-relay.csnet@RELAY.CS.NET yuasa%kurims.kurims.kyoto-u.junet%utokyo-relay.csnet@RELAY.CS.NET We have found them extremely responsive to bug reports and suggestions. SAMPLE TRANSCRIPT Below is a complete transcript for obtaining and installing KCL on a Sun-3. Make a directory for locating KCL tutorial% mkdir /usr/joe/kcl Get the compressed tar file tutorial% cd /usr/joe/kcl tutorial% ftp 128.83.144.1 220 rascal FTP server (Version 4.7 Sun Sep 14 12:44:57 PDT 1986) ready. Name: ftp Password: guest ftp>binary ftp>get /pub/kcl.tar.Z kcl.tar.Z ftp>quit Build the KCL directory structure tutorial% uncompress kcl.tar.Z tutorial% tar -xvf kcl.tar . tutorial% rm kcl.tar Make KCL tutorial% cd /usr/joe/kcl/ tutorial% su password: super-user-password tutorial# cp h/cmpinclude.h /usr/include tutorial# exit tutorial% make Edit and Install Two Files We wish to replace "~" by "/usr/joe" in lc and kcl, and put them in a directory on the search path e.g. "/usr/joe/bin" tutorial% cd /usr/joe/kcl/unixport tutorial% mkdir /usr/joe/bin tutorial% sed -e "s.~./usr/joe/kcl.g" lc > /usr/joe/bin/lc tutorial% sed -e "s.~./usr/joe/kcl.g" kcl > /usr/joe/bin/kcl tutorial% chmod a+x /usr/joe/bin/lc /usr/joe/bin/kcl It is now possible to run kcl: tutorial% /usr/joe/bin/kcl KCL (Kyoto Common Lisp) >(+ 2 3) 5 >(bye) It is best to become super user and execute the following commands, so that all users may execute kcl. tutorial% su Password: superuser-password tutorial# cp /usr/joe/bin/kcl /usr/local/bin tutorial# cp /usr/joe/bin/lc /usr/local/bin tutorial# exit This transcript puts the entire kcl system, including sources and documentation, in the directory /usr/joe/kcl. Any other directory name would do as well, e.g., /usr/local instead of /usr/joe. Although this transcript has worked perfectly for us on a Sun-3, it might not work for you if you are running under NFS but not logged into the right machine: you may need to login as root on the system where /usr/include and /usr/local/bin are really located to do the super-user things. Immediately after the make is finished about 8.4 megatyes of disk space are in use. SINCERELY YOURS Robert S. Boyer William F. Schelter cl.boyer@r20.utexas.edu atp.schelter@r20.utexas.edu This message was written by Robert S. Boyer and William F. Schelter. The opinions expressed are theirs and are not necessarily those of the authors of KCL, the University of Texas, or MCC. The authors of KCL have, however, indicated that they have no objection to our distributing this message. P.S. Thanks to Dave Capshaw, George Fann, Warren Hunt, Ken Rimey, and Carl Quillen for helping debug this message. Ken Rimey, rimey@ernie.Berkeley.EDU, makes the following remarks about bringing up this release of KCL under BSD 4.3 on a Vax. 1. Bringing up KCL under BSD4.3. The machine on which I installed kcl was a Vax 88xx running Ultrix V2.0. Kcl crashed when executing the final save-system command in init_kcl.lsp. It also did so on a Vax running BSD4.3. (I don't know of any Vaxen still running 4.2.) The problem is caused by some highly non-portable code introduced into Lsave() in c/unixsave.c since the version of March 28, 1986. I deleted the new code and reintroduced the old which had been commented out. Here is the resulting working Lsave(): Lsave() { char filename[256]; check_arg(1); check_type_or_pathname_string_symbol_stream(&vs_base[0]); coerce_to_filename(vs_base[0], filename); _cleanup(); memory_save(kcl_self, filename); exit(0); /* no return */ } KCL ran successfully after only fixing the Lsave problem. 2. The files o/makefile and unixport/makefile define variables that need to be changed when compiling for any machine other than a Sun-3. These definitions are found at the heads of these files. Here is the head of my copy of o/makefile: MACHINE = VAX # Select 'VAX', 'SUN', 'SUN2R3', 'SUN3', 'ISI', 'SEQ', 'IBMRT', # or 'NEWS'. CHTAB = char_table.s # Select 'char_table.s' or 'sun_chtab.s'. # 1) char_table.s : for VAX, SEQ and NEWS # 2) sun_chtab.s : for SUN, SUN2R3 and SUN3 # 3) isi_chtab.s : for ISI # 4) ibmrt_chtab.s: for IBMRT For machines other than Sun-3, one might change the MAKE KCL section of this message to: tutorial% cd /usr/joe/kcl/ tutorial% vi o/makefile (If not Sun-3, change definitions MACHINE and CHTAB.) tutorial% vi o/unixport (If not Sun-3, change definition of MACHINE.) tutorial% su password: super-user-password tutorial# cp h/cmpinclude.h /usr/include tutorial# exit tutorial% make ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jul 2 03:15:19 1987 Date: Thu, 2 Jul 87 03:15:09 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #163 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 2 Jul 87 03:09 EDT Received: from relay.cs.net by RELAY.CS.NET id aa25944; 1 Jul 87 8:39 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa07078; 1 Jul 87 8:38 EDT Date: Tue 30 Jun 1987 23:02-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #163 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Wednesday, 1 Jul 1987 Volume 5 : Issue 163 Today's Topics: Theory - The Symbol Grounding Problem & Graded Categories ---------------------------------------------------------------------- Date: 28 Jun 87 23:56:43 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem.... In article <919@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel writes: > ...... > > .... The feature extractor obviates the symbol-grounding > > problem. > > ..... You are vastly underestimating the problem of > sensory categorization, sensory learning, and the relation between > lower and higher-order categories. Nor is it obvious that symbol manipulation > can still be regarded as just symbol manipulation when the atomic symbols > are constrained to be the labels of sensory categories.... I still think we're having more trouble with terminology than we would have with the concepts if we understood each other. To get a little more concrete, how walking through what a machine might do in perceiving a chair? I was just looking at a kitchen chair, a brown wooden kitchen chair against a yellow wall, in side light from a window. Let's let a machine train its camera on that object. Now either it has a mechanical array of receptors and processors, like the layers of cells in a retina, or it does a functionally equivalent thing with sequential processing. What it has to do is compare the brightness of neighboring points to find places where there is contrast, find contrast in contiguous places so as to form an outline, and find closed outlines to form objects. There are some subtleties needed to find partly hidden objects, but I'll just assume they're solved. There may also be an interpretation of shadow gradations to perceive roundness. Now the machine has the outline of an object in 2 dimensions, and maybe some clues to the 3rd dimension. There are CAD programs that, given a complete description of an object in 3D, can draw any 2D view of it. How about reversing this essentially deductive process to inductively find a 3D form that would give rise to the 2D view the machine just saw. Let the machine guess that most of the odd angles in the 2D view are really right angles in 3D. Then, if the object is really unfamiliar, let the machine walk around the chair, or pick it up and turn it around, to refine its hypothesis. Now the machine has a form. If the form is still unfamiliar, let it ask, "What's that, Daddy?" Daddy says, "That's a chair." The machine files that information away. Next time it sees a similar form it says "Chair, Daddy, chair!" It still has to learn about upholstered chairs, but give it time. That brings me to a question: do you really want this machine to be so Totally Turing that it grows like a human, learns like a human, and not only learns new objects, but, like a human born at age zero, learns how to perceive objects? How much of its abilities do you want to have wired in, and how much learned? But back to the main question. I have skipped over a lot of detail, but I think the outline can in principle be filled in with technologies we can imagine even if we do not have them. How much agreement do we have with this scenario? What are the points of disagreement? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 29 Jun 87 08:49:00 EST From: cugini@icst-ecf.arpa Reply-to: Subject: invertibility as a graded category ? Harnad writes: > In responding to Cugini and Brilliant I misinterpreted a point that > the former had made and the latter reiterated. It's a point that's > come up before: What if the iconic representation -- the one that's > supposed to be invertible -- fails to preserve some objective property > of the sensory projection? ... The reply is that an analog > representation is only analog in what it preserves, not in what it fails > to preserve. Icons are hence approximate too. ... > There is no requirement that all the features of the sensory > projection be preserved in icons; just that some of them should be -- > enough to subserve our discrimination capacities. > ... But none of this > information loss in either sensory projections or icons (or, for that > matter, categorical representations) compromises groundedness. It just > means that our representations are doomed to be approximations. But then why say that icons, but not categorical representations or symbols are/must be invertible? (This was *your* original claim, after all)? Isn't it just a vacuous tautology to claim that icons are invertible wrt to the information they preserve, but not wrt the information they lose? How could it be otherwise? Aren't even symbols likewise invertible in that weak sense? (BTW, I quite agree that the information loss does not compromise grounding - indeed my very point was that there is nothing especially scandalous about non-invertible icons.) Look, there's information loss (many to one mapping) at each stage of the game: 1. distal object 2. sensory projection 3. icons 4. categorical representation 5. symbols It was you who seemed to claim that there was some special invertibility between stages 2 and 3 - but now you claim for it invertibility in only such a vitiated sense as to apply to all the stages. So a) do you still claim that the transition between 2 and 3 is invertible in some strong sense which would not be true of, say, [1 to 2] or [3 to 4], and b) if so, what is that sense? Perhaps you just want to say that the transition between 2 and 3 is usually more invertible than the other transitions ? John Cugini ------------------------------ Date: 29 Jun 87 10:35:00 EST From: cugini@icst-ecf.arpa Reply-to: Subject: epistemological exception-taking > > From me: > > > > What if there were a few-to-one transformation between the skin-level > > sensors ... > > My example was to suppose that #1: > > a combination of both red and green retinal receptors and #2 a yellow > > receptor BOTH generated the same iconic yellow. > From: Neil Hunt > > We humans see the world (to a first order at least) through red, green and > blue receptors. We are thus unable to distinguish between light of a yellow > frequency, and a mixture of light of red and green frequencies, and we assign > to them a single token - yellow. However, if our visual apparatus was > equipped with yellow receptors as well, then these two input stimuli > would *appear* quite different, as indeed they are. ... Oh, really? How do you claim to know what the mental effect would be of a hypothetical visual nerve apparatus? Do you know what it feels like to be a bat? John Cugini ------------------------------ Date: 29 Jun 87 20:53:28 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem: Against Rosch & Wittgenstein In article <931@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel asks: > > Why require 100% accuracy in all-or-none categorizing?... I learned > > recently that I can't categorize chairs with 100% accuracy. > > This is a misunderstanding. The "100% accuracy" refers to the > all-or-none-ness of the kinds of categories in question. The rival > theories in the Roschian tradition have claimed that many categories > (including "bird" and "chair") do not have "defining" features. Instead, > membership is either fuzzy or a matter of degree (i.e., percent).... OK: once I classify a thing as a chair, there are no two ways about it: it's a chair. But there can be a stage when I can't decide. I vacillate: "I think it's a chair." "Are you sure?" "No, I'm not sure, maybe it's a bed." I would never say seriously that I'm 40 percent sure it's a chair, 50 percent sure it's a bed, and 10% sure it's an unfamiliar object I've never seen before. I think this is in agreement with Harnad when he says: > Categorization preformance (with all-or-none categories) is highly reliable > (close to 100%) and MEMBERSHIP is 100%. Only speed/ease of categorization and > typicality ratings are a matter of degree.... > This is not to deny that even all-or-none categorization may encounter > regions of uncertainty. Since ALL category representations in my model are > provisional and approximate ..... it is always possible that > the categorizer will encounter an anomalous instance that he cannot classify > according to his current representation..... > ...... This still does not imply that membership is > fuzzy or a matter of degree..... So to pass the Total Turing Test, a machine should respond the way a human does when faced with inadequate or paradoxical sensory data: it should vacillate (or bluff, as some people do). In the presence of uncertainty it will not make self-consistent statements about uncertainty, but uncertain and possibly inconsistent statements about absolute membership. M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 29 Jun 87 23:34:58 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: The symbol grounding problem: "Fuzzy" categories? In comp.ai.digest: Laws@STRIPE.SRI.COM (Ken Laws) asks re. "Fuzzy Symbolism": > Is a mechanical rubber penguin a penguin?... dead...dismembered > genetically damaged or altered...? When does a penguin embryo become > a penguin?... I can't unambiguously define the class of penguins, so > how can I be 100% certain that every penguin is a bird?... and even > that could change if scientists someday discover incontrovertible > evidence that penguins are really fish. In short, every category is a > graded one except for those that we postulate to be exact as part of > their defining characteristics. I think you're raising the right questions, but favoring the wrong answers. My response to this argument for graded or "fuzzy" category was that our representations are provisional and approximate. They converge on the features that will reliably sort members from nonmembers on the basis of the sample of confusable alternatives encountered to date. Being always provisional and approximate, they are always susceptible to revision should the context of confusable alternatives be widened. But look at the (not so hidden) essentialism in Ken's query: "how can I be 100% certain that every penguin is a bird?". I never promised that! We're not talking about ontological essences here, about the way things "really are," from the God's Eye" or omniscient point of view! We're just talking about how organisms and other devices can sort and label APPEARANCES as accurately as they do, given the feedback and experiential sample they get. And this sorting and labeling is provisional, based on approximate representations that pick out features that reliably handle the confusable alternatives sampled to date. All science can do is tighten the approximation by widening the alternatives (experimentally) or strengthening the features (theoretically). But provisionally, we do alright, and it's NOT because we sort things as being what they are as a matter of degree. A penguin is 100% a bird (on current evidence) -- no more or less a bird than a sparrow. If tomorrow we find instances that make it better to sort and label them as fish, then tomorrow's approximation will be better than today's, but they'll then be 100% fish, and so on. Note that I'm not denying that there are graded categories; just that these aren't them. Examples of graded categories are: big, intelligent, beautiful, feminine, etc. > You are entitled to such an opinion, of course, but I do not > accept the position as proven... (Why opinion, by the way, rather than hypothesis, on the evidence and logical considerations available? Nor will this hypothesis be proven: just supported by further evidence and analysis, or else supplanted by a rival hypothesis that accounts for the evidence better; or the hypothesis and its supporting arguments may be shown to be incoherent or imparsimonious...) > ...We do, of course, sort and categorize objects when forced to do so. > At the point of observable behavior, then, some kind of noninvertible > or symbolic categorization has taken place. Such behavior, however, > is distinct from any of the internal representations that produce it. > I can carry fuzzy and even conflicting representations until -- and > often long after -- the behavior is initiated. Even at the instant of > commitment, my representations need be unambiguous only in the > implicit sense that one interpretation is momentarily stronger than > the other -- if, indeed, the choice is not made at random. I can't follow some of this. Categorization is the performance capacity under discussion here. ("Force" has nothing to do with it!). And however accurately and reliably people can actually categorize things, THAT'S how accurately our models must be able to do it under the same conditions. If there's successful all-or-none performance, the representational model must be able to generate it. How can the behavior be "distinct from" the representations that produce it? This is not to say that representations will always be coherent, or even that incoherent representations can't sometimes generate correct categorization (up to a point). But I hardly think that the basis of the bulk of our reliable all-or-none sorting and labeling will turn out to be just a matter of momentary relative strengths -- or even chance -- among graded representations. I think probabilistic mechanisms are more likely to be involved in feature-finding in the training phase (category learning) rather than in the steady state phase, when a (provisional) performance asymptote has been reached. > It may also be true that I do reduce some representations to a single > neural firing or to some other unambiguous event -- e.g., when storing > a memory. I find this unlikely as a general model. Coarse coding, > graded or frequency encodings, and widespread activation seem better > models of what's going on. Symbolic reasoning exists in pure form > only on the printed page; our mental manipulation even of abstract > symbols is carried out with fuzzy reasoning apparatus. Some of this sounds like implementational considerations rather than representational ones. The question was: Do all-or-none categories (such as "bird") have "defining" features that can be used to sort members from nonmembers at the level of accuracy (~100%) with which we sort? However they are coded, I claim that those features MUST exist in the inputs and must be detected and used by the categorizer. A penguin is not a bird as a matter of degree, and the features that reliably assign it to "bird" are not graded. Nor is "bird" a fuzzy category such as "birdlike." And, yes, symbolic representations are likely to be more apodictic (i.e., categorical) than nonsymbolic ones. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ End of AIList Digest ******************** From in%@vtcs1 Thu Jul 2 03:15:03 1987 Date: Thu, 2 Jul 87 03:14:54 edt From: in%AIList@stripe.sri.com@vtcs1 To: ailist@stripe.sri.com Subject: AIList Digest V5 #164 Status: R Received: from relay.cs.net by vtcs1.cs.vt.edu; Thu, 2 Jul 87 03:06 EDT Received: from relay.cs.net by RELAY.CS.NET id aa24605; 1 Jul 87 3:39 EDT Received: from stripe.sri.com by RELAY.CS.NET id aa05269; 1 Jul 87 3:36 EDT Date: Tue 30 Jun 1987 23:07-PDT From: AIList Moderator Kenneth Laws Subject: AIList Digest V5 #164 To: AIList@stripe.sri.com Reply-to: AIList@stripe.sri.com US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 AIList Digest Wednesday, 1 Jul 1987 Volume 5 : Issue 164 Today's Topics: Theory - The Symbol Grounding Problem ---------------------------------------------------------------------- Date: 30 Jun 87 00:19:12 GMT From: mind!harnad@princeton.edu (Stevan Harnad) Subject: Re: The symbol grounding problem marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel asks: > how about walking through what a machine might do in perceiving a chair? > ...let a machine train its camera on that object. Now either it > has a mechanical array of receptors and processors, like the layers > of cells in a retina, or it does a functionally equivalent thing with > sequential processing. What it has to do is compare the brightness of > neighboring points to find places where there is contrast, find > contrast in contiguous places so as to form an outline, and find > closed outlines to form objects... Now the machine has the outline > of an object in 2 dimensions, and maybe some clues to the 3rd > dimension... inductively find a 3D form that would give rise to the > 2D view the machine just saw... Then, if the object is really > unfamiliar, let the machine walk around the chair, or pick it > up and turn it around, to refine its hypothesis. So far, apart from its understandable toward current engineering hardware concepts, there is no particular objection to this description of a stereoptic sensory receptor. > Now the machine has a form. If the form is still unfamiliar, > let it ask, "What's that, Daddy?" Daddy says, "That's a chair." > The machine files that information away. Next time it sees a > similar form it says "Chair, Daddy, chair!" It still has to > learn about upholstered chairs, but give it time. Now you've lost me completely. Having acknowledged the intricacies of sensory transduction, you seem to think that the problem of categorization is just a matter of filing information away and finding "similar forms." > do you really want this machine to be so Totally Turing that it > grows like a human, learns like a human, and not only learns new > objects, but, like a human born at age zero, learns how to perceive > objects? How much of its abilities do you want to have wired in, > and how much learned? That's an empirical question. All it needs to do is pass the Total turing Test -- i.e., exhibit performance capacities that are indistinguishable from ours. If you can do it by building everything in a priori, go ahead. I'm betting it'll need to learn -- or be able to learn -- a lot. > But back to the main question. I have skipped over a lot of > detail, but I think the outline can in principle be filled in > with technologies we can imagine even if we do not have them. > How much agreement do we have with this scenario? What are > the points of disagreement? I think the main details are missing, such as how the successful categorization is accomplished. Your account also sounds as if it expects innate feature detectors to pick out objects for free, more or less nonproblematically, and then serve as a front end for another device (possibly a conventional symbol-cruncher a la standard AI?) that will then do the cognitive heavy work. I think that the cognitive heavy work begins with picking out objects, i.e., with categorization. I think this is done nonsymbolically, on the sensory traces, and that it involves learning and pattern recognition -- both sophisticated cognitive activities. I also do not think this work ends, to be taken over by another kind of work: symbolic processing. I think that ALL of cognition can be seen as categorization. It begins nonsymbolically, with sensory features used to sort objects according to their names on the basis of category learning; then further sorting proceeds by symbolic descriptions, based on combinations of those atomic names. This hybrid nonsymbolic/symbolic categorizer is what we are; not a pair of modules, one that picks out objects and the other that thinks and talks about them. -- Stevan Harnad (609) - 921 7771 {bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad harnad%mind@princeton.csnet harnad@mind.Princeton.EDU ------------------------------ Date: 30 Jun 87 20:52:21 GMT From: diamond.bbn.com!aweinste@husc6.harvard.edu (Anders Weinstein) Subject: Re: The symbol grounding problem In reply to my statement that >> the *semantic* meaning of a symbol is still left largely unconstrained >> even after you take account of it's "grounding" in perceptual >> categorization. This is because what matters for intentional content >> is not the objective property in the world that's being detected, but >> rather how the subject *conceives* of that external property, a far >> more slippery notion... Stevan Harnad (harnad@mind.UUCP) writes: > > As to what people "conceive" themselves to be categorizing: My model > is proposed in a framework of methodological epiphenomenalism. I'm > interested in what's going on in people's heads only inasmuch as it is > REALLY generating their performance, not just because they think or > feel it is. I regret the subjectivistic tone of my loose characterization; what people can introspect is indeed not at issue. I was merely pointing out that the *meaning* of a symbol is crucially dependent on the rest of the cognitive system, as shown in the Churchland's example: >> ... primitive people may be able to reliably >> categorize certain large-scale atmospheric electrical discharges; >> nevertheless, the semantic content of their corresponding states might >> be "Angry gods nearby" or some such. >> > ... "Angry gods nearby" is not just an atomic label for > "thunder" (otherwise it WOULD be equivalent to it in my model -- both > labels would pick out approximately the same thing); in fact, it is > decomposable, and hence has a different meaning in virtue of the > meanings of "angry" and "gods." There should be corresponding internal > representational differences (iconic, categorical and symbolic) that > capture that difference. "Angry gods nearby" is composite in *English*, but it need not be composite in native, or, more to the point, in the supposed inner language of the native's categorical mechanisms. They may have a single word, say "gog", which we would want to translate as "god-noise" or some such. Perhaps they train their children to detect gog in precisely the same way we train children to detect thunder -- our internal thunder-detectors are identical. Nevertheless, the output of their thunder-detector does not *mean* "thunder". Let me try to clarify the point of these considerations. I am all for an inquiry into the mechanisms underlying our categorization ablities. Anything you can discover out about these mechanisms would certainly be a major contribution to psychology. My only concern is with semantics: I was piqued by what seemed to be an ambitious claim about the significance of the psychology of categorization for the problem of "intentionality" or intrinsic meaningfulness. I merely want to emphasize that the former, interesting though it is, hardly makes a dent in the latter. As I said, there are two reasons why meaning resists explication by this kind of psychology: (1) holism: the meaning of even a "grounded" symbol will still depend on the rest of the cognitive system; and (2) normativity: meaning is dependent upon a determination of what is a *correct* response, and you can't simply read such a norm off from a description of how the mechanism in fact performs. I think these points, particularly (1), should be quite clear. The fact that a subject's brain reliably asserts the symbol "foo" when and only when thunder is presented in no way "fixes" the meaning of "foo". Of course it is obviously a *constraint* on what "foo" may mean: it is in fact part of what Quine called the "stimulus meaning" of "foo", his first constraint on acceptable translation. Nevertheless, by itself it is still way too weak to do the whole job, for in different contexts the postive output of a reliable thunder-detector could mean "thunder", something co-extensive but non-synonymous with "thunder", "god-noise", or just about anything else. Indeed, it might not *mean* anything at all, if it were only part of a mechanical thunder-detector which couldn't do anything else. I wonder if you disagree with this? As to normativity, the force of problem (2) is particularly acute when talking about the supposed intentionality of animals, since there aren't any obvious linguistic or intellectual norms that they are trying to adhere to. Although the mechanics of a frog's prey-detector may be crystal clear, I am convinced that we could easily get into an endless debate about what, if anything, the output of this detector really *means*. The normativity problem is germane in an interesting way to the problem of human meanings as well. Note, for example, that in doing this sort of psychology, we probably won't care about the difference between correctly identifying a duck and mis-identifying a good decoy -- we're interested in the perceptual mechanisms that are the same in both cases. In effect, we are limiting our notion of "categorization" to something like "quick and largely automatic classification by observation alone". We pretty much *have* to restrict ourselves in this way, because, in the general case, there's just no limit to the amount of cognitive activity that might be required in order to positively classify something. Consider what might go into deciding whether a dolphin ought to be classified as a fish, whether a fetus ought to be classified as a person, etc. These decisions potentially call for the full range of science and philosophy, and a psychology which tries to encompass such decisions has just bitten off more than it can chew: it would have to provide a comprehensive theory of rationality, and such an ambitious theory has eluded philosophers for some time now. In short, we have to ignore some normative distinctions if we are to circumscribe the area of inquiry to a theoretically tractable domain of cognitive activity. (Indeed, in spite of some of your claims, we seem committed to the notion that we are limiting ourselves to particular *modules* as explained in Fodor's modularity book.) Unfortunately -- and here's the rub -- these normative distinctions *are* significant for the *meaning* of symbols. ("Duck" doesn't *mean* the same thing as "decoy"). It seems that, ultimately, the notion of *meaning* is intimately tied to standards of rationality that cannot easily be reduced to simple features of a cognitive mechanism. And this seems to be a deep reason why a descriptive psychology of categorization barely touches the problem of intentionality. Anders Weinstein BBN Labs ------------------------------ Date: 30 Jun 87 19:02:28 GMT From: teknowledge-vaxc!dgordon@beaver.cs.washington.edu (Dan Gordon) Subject: Re: The symbol grounding problem: Against Rosch & Wittgenstein In article <931@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes: >(And I must repeat: Whether or not we can introspectvely report the features >we are actually using is irrelevant. As long as reliable, consensual, >all-or-none categorization performance is going on, there must be a set of >underlying features governing it -- both with sensory and more Is this so? There is no reliable, consensual all-or-none categorization performance without a set of underlying features? That sounds like a restatement of the categorization theorist's credo rather than a thing that is so. Dan Gordon ------------------------------ Date: 30 Jun 87 20:49:32 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <937@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > ... > marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel asks: > > how about walking through what a machine might do in perceiving a chair? > > ... (a few steps skipped here) > > Now the machine has a form. If the form is still unfamiliar, > > let it ask, "What's that, Daddy?" Daddy says, "That's a chair." > > The machine files that information away. Next time it sees a > > similar form it says "Chair, Daddy, chair!" ... > > Now you've lost me completely. Having acknowledged the intricacies of > sensory transduction, you seem to think that the problem of categorization > is just a matter of filing information away and finding "similar forms." I think it is. We've found a set of lines, described in 3 dimensions, that can be rotated to match the outline we derived from the view of a real chair. We file it in association with the name "chair." A "similar form" is some other outline that can be matched (to within some fraction of its size) by rotating the same 3D description. > I think the main details are missing, such as how the successful > categorization is accomplished...... Are we having a problem with the word "categorization"? Is it the process of picking discrete objects out of a pattern of light and shade ("that's a thing"), or the process of naming the object ("that thing is a chair")? > ..... Your account also sounds as if it > expects innate feature detectors to pick out objects for free, more or > less nonproblematically..... You left out the part where I referred to computer-aided-design modules. I think we can find outlines by looking for contiguous contrasts. If the outlines are straight we (the machine, maybe also humans) can define the ends of the straight lines in the visual plane, and hypothesize corresponding lines in space. If hard-coding this capability gives an "innate feature detector" then that's what I want. > ...... and then serve as a front end for another > device (possibly a conventional symbol-cruncher a la standard AI?) > that will then do the cognitive heavy work. I think that the cognitive > heavy work begins with picking out objects, i.e., with categorization. I think I find objects with no conscious knowledge of how I do it (is that what you call "categorization")? Saying what kind of object it is more often involves conscious symbol-processing (sometimes one forgets the word and calls a perfectly familiar object "that thing"). > I think this is done nonsymbolically, on the sensory traces, and that it > involves learning and pattern recognition -- both sophisticated > cognitive activities. If you're talking about finding objects in a field of light and shade, I agree that it is done nonsymbolically, and everything else you just said. > ..... I also do not think this work ends, to be taken > over by another kind of work: symbolic processing..... That's where I have trouble. Calling a penguin a bird seems to me purely symbolic, just as calling a tomato a vegetable in one context, and a fruit in another, is a symbolic process. > ..... I think that ALL of > cognition can be seen as categorization. It begins nonsymbolically, > with sensory features used to sort objects according to their names on > the basis of category learning; then further sorting proceeds by symbolic > descriptions, based on combinations of those atomic names. This hybrid > nonsymbolic/symbolic categorizer is what we are; not a pair of modules, > one that picks out objects and the other that thinks and talks about them. Now I don't understand what you said. If it begins nonsymbolically, and proceeds symbolically, why can't it be done by linking a nonsymbolic module to a symbolic module? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ Date: 30 Jun 87 19:47:08 GMT From: ihnp4!homxb!houdi!marty1@ucbvax.Berkeley.EDU (M.BRILLIANT) Subject: Re: The symbol grounding problem In article <937@mind.UUCP>, harnad@mind.UUCP (Stevan Harnad) writes: > marty1@houdi.UUCP (M.BRILLIANT) of AT&T Bell Laboratories, Holmdel asks: > .... > > do you really want this machine to be so Totally Turing that it > > grows like a human, learns like a human, and not only learns new > > objects, but, like a human born at age zero, learns how to perceive > > objects? How much of its abilities do you want to have wired in, > > and how much learned? > > That's an empirical question. All it needs to do is pass the Total > turing Test -- i.e., exhibit performance capacities that are > indistinguishable from ours. If you can do it by building everything > in a priori, go ahead. I'm betting it'll need to learn -- or be able to > learn -- a lot. To refine the question: how long do you imagine the Total Turing Test will last? Science fiction stories have robots or aliens living in human society as humans for periods of years, as long as they live with strangers, but failing after a few hours trying to supplant a human and fool his or her spouse. By "performance capabilities," do you mean the capability to adapt as a human does to the experiences of a lifetime? Or only enough learning capability to pass a job interview? M. B. Brilliant Marty AT&T-BL HO 3D-520 (201)-949-1858 Holmdel, NJ 07733 ihnp4!houdi!marty1 ------------------------------ End of AIList Digest ********************