From comsat@vpics1 Thu Oct 10 22:21:10 1985 Date: Thu, 10 Oct 85 22:21:06 edt From: comsat@vpics1.VPI To: fox@opus (FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: R Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a006708; 10 Oct 85 1:37 EDT Date: Wed 9 Oct 1985 21:13-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #140 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Thu, 10 Oct 85 22:05 EST AIList Digest Thursday, 10 Oct 1985 Volume 3 : Issue 140 Today's Topics: Query - Prolog on Macintoshes and IBM-Style PCs, AI Tools - Prolog vs. Lisp ---------------------------------------------------------------------- Date: Wed, 9 Oct 85 17:38:09 mdt From: ted%nmsu.csnet@CSNET-RELAY.ARPA Subject: prolog's on macintoshes and ibm-style pc's Currently I know about arity prolog and prolog-2. These are reasonably-featured prolog interpreter/compiler packages that run on the ibm pc. Does anyone know of other packages which are available on the pc or the macintosh??? ------------------------------ Date: 05 Sep 85 18:29:55 PDT (Thu) From: Sanjai Narain Subject: Response to Hewitt [Forwarded from the Prolog Digest by Laws@SRI-AI.] >Prolog (like APL before it) will fail as the foundation for >Artificial Intelligence because of competition with Lisp. >There are commercially viable Prolog implementations written >in Lisp but not conversely. For the same reason, Lisp should have failed as a foundation for computing because of competition with assembly language. There are commercially viable implementations of Lisp in assembly language but not conversely. >LOGIC as a PROGRAMMING Language will fail as the foundation >for AI because: >1. Logical inference cannot be used to infer the decisions > that need to be taken in open systems because the decisions > are not determined by system inputs. >>2. Logic does not cope well with the contradictory knowledge >> bases inherent in open systems. It leaves out >> counterarguments and debate. >>3. Taking action does not fit within the logic paradigm. 1. Hewitt clearly states in his recent BYTE article that traditional notions of computation as defined, for example, by Turing machines or recursive functions cannot model the behavior of open systems. Hence even Lisp is inadequate for such modeling (by his reasoning). 2. The notion of contradiction (i.e. inconsistency) is well understood in logic. 3. The statement is too vague for debate. What do the words "action" and "fit" mean? Certainly, if action can be modeled by an effective procedure, it can be modeled by logic, cf. 1. -- Sanjai Narain ------------------------------ Date: Thu, 5-Sep-85 13:40:43 PDT From: (Tom Khabaza) mcvax!ukc!warwick!cvaxa!tomk@Seismo Subject: On Hewitt's "Prolog and logic programming will fail" [Forwarded from the Prolog Digest by Laws@SRI-AI.] I have read with interest the discussion following Carl Hewitt's "Prolog will fail as the foundation for AI and so will Logic Programming". I particularly enjoyed Vijay Saraswat's reply, most of which I agree with. However, I would like to add a few comments: In some ways I was surprised by the original message; I should have thought that if AI has taught us anything, it is that to solve a given problem, we need a good representation language. Why anyone might think that logic is the BEST representation language for every problem is beyond me. (No Kowalskiist flames please, I know the arguments, and I don't regard the case as proven.) On the other hand, we don't yet know what the limits of logic programming are; researchers in the field are constantly coming up with new techniques. There is convincing evidence that logic programming is better than conventional programming for some kinds of task, at least with regard to ease and clarity (though probably not yet efficiency). But I think the basis of the original comment goes deeper than the virtues and vices of logic programming. As I understand it (and I wasn't around at the time) some earlier AI programming languages, such as perhaps micro-Planner and its successors, WERE expected to become a "foundation" for AI. Perhaps this was because people still had hopes for the notion of some "ultimate" representation language, or family of languages. AI is older and perhaps more cynical now; I don't think we expect some single foundation to the field in the form of a representation language. Logic programming may be very useful for some parts of AI; for example some kinds of rule based systems, but I don't expect it to be the best tool for all kinds of AI programming. In fact my personal opinion is that logic programming will find its forte in more conventional Computer Science, where formal specification is a more practical proposition than in the relatively exploratory activity of AI programming. But I will say this in its favour: logic programming is IMPORTANT. Logic programming is as different from conventional programming as programming is from not programming at all. I have met people who have given up on Prolog because it was difficult for them and they (rightfully) considered themselves competent programmers - and so thought it must be Prolog's fault! (I don't mean to imply that anyone who has posted in this discussion is such a person.) But logic programming is different in fundamental ways; it's worth presevering to get to the bottom of it, and as logic programming languages improve, it will become even more so. So for all you computer people out there, USE Prolog, and study how other people have used it. It really is worth it. ------------------------------ Date: Fri, 13 Sep 85 10:41:27 bst From: William Clocksin Subject: Lisp/Prolog [Forwarded from the Prolog Digest by Laws@SRI-AI.] I received issues 36 and 37 late owing to netproblem somewhere between Stanford and Cambridge. I am puzzled by this Lisp/Prolog debate started by Carl Hewitt. I use both Prolog and Lisp, and have never felt the need to use one exclusively. I suppose they are like screwdrivers and chisels; both are roughly the same, but for slightly different purposes; to a person unfamiliar with one of them, the other one might seem redundant. I am also puzzled about the question of a "foundation" for AI. How can a language be a "foundation" for anything? Was Latin a "foundation" of Western civilisation? Seen any fundamental native speakers lately? Besides, does AI deserve to have foundations attributed to it anyway? Another problem is this question about logic. Prolog is a programming language. It was inspired by logic, but it is not programming in logic. Proponents of using logic do have a problem matching impedance with the real world. But Prolog is to logic as Lisp is to lambda calculus. Those who advocate programming in lambda calculus have the same problem as those who advocate programming in pure logic. If Prolog can be said to have any connection with logic, it is as the FORTRAN of logic programming. Prolog is useful because you can grow data structures that have actual variables in them, and because it is easy to define nondeterministic methods. I know how Prolog searches for a solution just as I know how flow of control happens in Lisp, say. I am not disappointed with Prolog's strict strategy just as I am not disappointed with Lisp's inability to run programs backwards, say. I take it as it comes, and it is useful for some things. Talking hypothetically about the "ideal" language is another topic entirely, and it only muddies the water to bring Prolog and Lisp into it. ------------------------------ Date: Wed 25 Sep 85 11:22:32-PDT From: Fernando Pereira Subject: Prolog vs. Lisp (again!) [Forwarded from the Prolog Digest by Laws@SRI-AI.] I humbly confess my incompetence at the debating style Carl Hewitt is using in the Prolog vs. Lisp discussions, which seems to consist in ignoring the FACTUAL points of other contributions and just continuing to repeat the same OPINIONS. It is a FACT that no practical Prolog system is written entirely in Lisp: Common, Inter or any other. Fast Prolog systems have been written for Lisp machines (Symbolics, Xerox, LMI) but their performance depends crucially on major microcode support (so much so that the Symbolics implementation, for example, requires additional microstore hardware to run Prolog). The reason for this is simple: No Lisp (nor C, for that matter...) provides the low-level tagged-pointer and stack operations that are critical to Prolog performance. The fact that Lisp is not good enough as an implementation language for Prolog should not be considered as a weakness of Lisp, BECAUSE Lisp was not designed for such low-level operations in the first place. In fact, NO ``high-level'' programming language that I know of provides those kinds of operations, and for the very simple reason that, being high-level languages, they have no business in exploring the recesses of particular machine architectures. ALL fast Prolog systems that I have seen (some of which I helped implement) rely on a careful exploitation of the underlying machine architecture. By the same argument, the fact that Lisp cannot be efficiently implemented in Prolog cannot be the basis of a valid criticism of Prolog. Prolog is not a systems programming language, and in any case a good Lisp implementation must be carefully coupled to the underlying machine architecture -- so much so the the fastest Lisps rely on specialized architectures! It seems clear to me that no single existing programming language can be said to provide a ``foundation'' for AI. In fact, the very notion of a programming language providing a foundation for a scientific subject seems to me rather misguided. Does Fortran provide a ``foundation'' for physics? The relation between AI problems, formal descriptions and programming concepts is far too subtle for us to expect a ``foundation'' for AI in a mere programming language. The crusading tone of Hewitt's comments is also rather unsettling. AI researchers will use whatever language they feel most comfortable with for the problem they are working on, without need for any guidance from on high as to the ultimate suitability of that language. If more researchers use Prolog, is that a threat to Lisp users? If I do a piece of AI research using Prolog, will it not be judged according to its content, independently of programming language? That kind of battle might be very important for AI software companies, but surely we should not let marketing hype get in the way of our research. I am sitting at a Sun workstation typing this, with a Prolog window just to the right. Will my research be useless to someone who sits at a Xerox 1109? If I walk down the corridor and write a Lisp program on a Symbolics machine (as I have done and surely will continue to do), will THAT work have a different value? If I decide to use Hoare's CSP for the navigation component of our robot, will I be then outside AI, because I am not using an ``official'' AI language? With respect to Rick McGeer's points: there are some elegant ways of compiling Prolog into Lisp, particularly into those statically scoped variety (or into other statically-scoped languages such as Algol-68...). I have reason to believe that a compiler along these lines would produce code considerably faster than the 100 LIPS he reports, even though still much slower than what is attainable with a lower-level implementation. [...] -- Fernando Pereira ------------------------------ Date: Thu, 26 Sep 85 10:29 EDT From: Tim Finin Subject: Lisp and Prolog [Forwarded from the Prolog Digest by Laws@SRI-AI.] PROLOG/LISP = LISP/C I'd like to amplify a point of view Clocksin put forth in V3 #39 in the great LISP vs. PROLOG debate. Prolog (and Logic Programming in general) and Lisp are both tools which are suited for different tasks. Luckily, none of us is being forced to use one to the exclusion of the other. I've had to give more than my share of introductory AI talks over the years. When the discussion gets around to Lisp I usually point out that Lisp is especially attractive for experimental programming situations, i.e. where you know what you want to accomplish but do not yet have all of the details as to algorithms, data structures, etc. worked out. Once you've worked out the last detail, you can re-implement your system in C, if you like, and gain the benefits of a faster and smaller system. Along these lines, I think that the slogan "Prolog is to Lisp as Lisp is to C" is not too inaccurate. I think that Prolog is even better suited to initial, experimental and exploratory attempts to attack a problem computationally. I find that it is a very convienent paradigm in which to get started. Once I have a better idea of how to reprent a problem and how to manipulate the representation, I can re-implement it in Lisp and gain a faster more steamlined system. -- Tim ------------------------------ Date: 9 Oct 85 12:56 PDT From: Kahn.pa@Xerox.ARPA Subject: Prolog vs Lisp Apropos the long debate started by Carl Hewitt: As co-author of LM-Prolog, the best performing Prolog implemented in Lisp, I thought some performance numbers might be useful. Naive reverse runs at 10K Lips on a CADR Lisp Machine with special micro-code support, primarily for unification and trailing. Without any micro-code support it runs 2 to 3 times slower. While there are much faster Prologs available, I would argue that LM-Prolog is commercially viable without the micro-code. There have been sales of the 3600 version of LM-Prolog despite the fact that it is not supported by micro-code. But part of Fernando's counter to Carl was that a serious Prolog implementation needs sub-primitives that Lisp does not provide. It is true that LM-Prolog even without micro-code relies on Zeta-Lisp's sub-primitives to manipulate pointers and create invisible pointers. This makes de-referencing variables very fast. While I don't have any figures I don't think that is so important, at least for the naive reverse benchmark. In other words I believe a pure Common Lisp implementation of Prolog on say a 3600 would run 3 or 4 times slower than Symbolics Prolog (which is fully micro-coded). Depending upon how important a factor of 3 or 4 is, one evaluates differently Carl's claim that Lisp is good for implementing Prolog (and not vice versa). I think part of this whole debate is confused with the larger debate of single paradigm vs multi-paradigm languages. My feeling is that while a single paradigm system is elegant that too often it doesn't fit the problem well and ackward cliches are used. For example, it is widely believed that for some kinds of problems object-oriented programming is most appropriate because it encasulates state and behavior so well. Concurrent Prolog advocates in such situations program objects in a complex cliche of tail recursive predicates where one argument is a stream of messages. No serious object-oriented language requires that each method list all the instance variables in the head and their new values again at the end of each "method" (the tail recursive call). I am not happy with the argument that goes -- well some problems are best programmed with Lisp, others with Prolog, others with SmallTalk, and still others with Ops5. Any significantly large problem is going to have sub-problems that are best handled by different paradigms. The debate should not be Lisp vs. Prolog but how can we combine Lisp and Prolog (and Smalltalk and ...) in a coherent well-integrated fashion. Its not easy. LM-Prolog was one attempt at doing this, as well as ICOT's ESP, Prolog-KR and LogLisp. I tried to integrate Prolog with Loops. None of these integrations are perfect but I think this is the direction to go for BUILDING TOOLS for BUILDING REAL APPLICATIONS. The CommonLoops effort at Xerox represents to me the best effort to date to build a tight integration of two paradigms (object and procedure-oriented). In contrast, to what I just said I think the single paradigm approach can be a great research strategy. Much of the Logic Programming community is caught up in the game of finding out how far can one go with logic programming. Can one write simulators, text editors, graphics, operating systems, embedded languages, and so on in Prolog or a language like it? It is rightfully considered cheating to "escape to Lisp" or jump into some object-oriented subsystem. Their purpose is to explore the paradigm itself -- its uses, its limitations, to stretch it and pull it in new directions not to build real applications. When building real applications the question is not can this or that be done in Prolog, we all know that everything can be written in Prolog, but what language can give the best support for building the application in the most fitting way. ------------------------------ End of AIList Digest ******************** From LAWS@KL.SRI.COM Fri Dec 18 03:46:52 1987 Mail-From: LAWS created at 9-Oct-85 21:30:41 Date: Wed 9 Oct 1985 21:20-PDT From: AIList Moderator Kenneth Laws Reply-To: AIList@SRI-AI Us-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #141 To: AIList@SRI-AI Resent-Date: Fri 18 Dec 87 00:08:32-PST Resent-From: Ken Laws Resent-To: isr@vtopus.CS.VT.EDU Resent-Message-Id: <12359392102.20.LAWS@KL.SRI.COM> Status: R AIList Digest Thursday, 10 Oct 1985 Volume 3 : Issue 141 Today's Topics: Psychology & Logic - Counterexample to Modus Ponens, AI Tools - DO you really need an AI machine? ---------------------------------------------------------------------- Date: Mon, 7 Oct 85 14:07 EDT From: Stephen G. Rowley Subject: Formal Logic Date: Wed, 2 Oct 85 14:11:34 edt From: John McLean Subject: Re: A Counterexample to Modus Ponens (1) If a Republican wins the election then if the winner is not Ronald Reagan, then the winner will be John Anderson. (1a) If a Republican wins the election and the winner is not Ronald Reagan, then the winner will be John Anderson. I would like to see some further discussion of this since I'm afraid that I don't see the difference between (1) and (1a) either. Certainly there is no difference with respect to inferential power as far as classical logic is concerned. There's no difference. (At least, not logically. Interpreting people's judgements about probabilities from natural language statements is an extremely subtle art, as I believe Chris said in a reply.) [Notation: -> means "implies", ~ means "not", & means "and", | means "inclusive or".] Let R = "a Republican wins the election" RR = "Ronald Reagan wins the election" JA = "John Anderson" wins the election" [1] R -> [ ~RR -> JA ] [1a] [ R & ~RR ] -> JA Of course, p -> q is the same as ~p | q. Then [1] and [1a] both transform to ~R | RR | JA ------------------------------ Date: Mon, 7 Oct 85 17:39:03 edt From: pugh@GVAX.CS.CORNELL.EDU (William Pugh) Subject: Re: A Counterexample to Modus Ponens >> In a recent issue of AIList, John McLean cited an article about >> inconsistencies in public opinion that apparently said: >> >> Before the 1980 presidential election, many held the two beliefs >> below: >> >> (1) If a Republican wins the election then if the winner is not >> Ronald Reagan, then the winner will be John Anderson. >> >> (2) A Republican will win the election. >> >> Yet few if any of these people believed the conclusion: >> >> (3) If the winner is not Reagan then the winner will be Anderson. >> Just to throw my two bits in: First, from probability, some background: P(x) is the probability of x & P(x|y) is the probability of x given that y is true. And by Bayes' theorem, P(x) P(y|x) P(x|y) = ----------- P(y) The beliefs stated above can be rephrased as follows: Let RW stand for "A Republican wins" Let RR stand for "Ronald Reagan wins" Let JA stand for "John Anderson wins" (1) P( ~RR => JA | RW) = 1 (2) P(RW) = 0.96 (for the sake of argument - high at any rate) and we wish to find (3) P(JA|~RR) well, from Bayes' theorem, P(JA) P(~RR|JA) P(JA) P(JA|~RR) = --------------- = ------- P(~RR) P(~RR) For example, if P(JA) = 0.01 and P(RR) = 0.95, then P(JA|~RR) = 0.2 Note that we have not used 1 at all, and made the (obvious) assumption (similar to 1) that if Anderson wins, Reagan can not win. Now, to try to figure out what went wrong in the original example, consider: Given P(A|B) and P(B), P(A&B) = P(A|B)P(B) HOWEVER, P(A&B) <= P(A) since P(A) = P(A&B)/P(B|A) THEREFORE, if A=>B with high probability and A with high probability THEN, B with high probability Now, going back to our example, we see the the original conclusion: >> (3) If the winner is not Reagan then the winner will be Anderson. is almost certainly true. Although non-obvious, this is because (3) is true if Reagan wins. The problem, therefore, is that people do not use the "standard" definition of implication. By "if A then B" people tend to think "given that A is true, B is true" - if A is false, the validity of the statement is not verified one way or the other. You can find more on Bayesian Inference in "Introduction to Artificial Intelligence" by Charniak and McDermott, or in many other sources. I have not yet figured out how to make Bayesian Inference work with this style of implication, but it is obvious that it requires some form of special treatment. I'll let you know if I figure anything out. Bill Pugh Cornell University ..{uw-beaver|vax135}!cornell!pugh 607-256-4934,ext5 ------------------------------ Date: Tue, 8 Oct 85 13:03 EDT From: Carole D Hafner Subject: More on Modus Ponens Several ideas have been proposed to explain the fact that many people in 1980 would agree to the following: 1. If a Republican wins the election then if it is not RR then it will be JA. 2. A Republican will win the election. But they would not agree to the apparent logical consequence: 3. If RR does not win the election then JA will win. Does this mean that modus ponens is not a rule of common sense reasoning? NO. The problem is due to the fact that "a republican" in the first sentence has an "intensional" meaning, while in the second it had (to most people) an extensional meaning. In other words, people believed: (there-exists X) [win(X,election) & party(X,republican)] and not (forall X) [win(X,election) --> party(X,republican)] It is the second interpretation of the indefinite noun phrase that gives rise to the conclusion in (3). Carole Hafner hafner@northeastern ------------------------------ Date: Tue 8 Oct 85 10:26:00-PDT From: EDWARDS@SRI-AI.ARPA Subject: Equivocation (?) in "failure" of modus ponens As far as I have been able to determine, in conversation with Todd Davies and Marcel Schoppers, the apparent failure of modus ponens rests on a subtle point about the understanding of "if-then". If all conditionals are taken as truth-functional, then most people would have believed the premises *and* the conclusion: (1) If a Republican wins then if Reagan doesn't win then Anderson will win. (2) A Republican will win. Therefore, (3) if Reagan doesn't win then Anderson will win. For (3), when read truth-functionally, is equivalent to: (4) Either Reagan will win or Anderson will win which in turn follows truth-functionally from: (5) Reagan will win which most people believed. The problem is due to the fact that (3) is not normally understood truth-functionally; it is understood as a counterfactual, setting up a possible situation (or "mental space"--Fauconnier, *Mental Spaces*) in which Reagan doesn't win and asking what assumption will make that situation most like the actual one. The assumption most people would make, given such a situation, is that the Democrats will win; so they believed (6), not (3): (6) If Reagan doesn't win, then the Democrats will win. The really interesting question here is whether (1) and (2) are understood truth-functionally. If they were, then the alleged failure of modus ponens would rest on a simple equivocation. But I don't think they are. (1) is a counterfactual just like (3) and (6). The problem is that (3) is understood quite differently when it appears as the consequent of (1) than when it appears alone. The antecedent of (1) sets up a mental space in which a Republican wins, by hypothesis. This affects the understanding of (3) in a way in which a mere factual belief that a Republican will win (such as is expressed in (2)) does not. It rules out the consideration of a Democratic victory even in a counterfactual situation. So when the antecedent of (3) sets up another mental space inside the first--where it is presupposed that Reagan doesn't win--the consequent of (6) is ruled out. Inside the second mental space, *by hypothesis*, a Republican wins but Reagan doesn't win. Thus, Anderson's victory is the only conclusion available. Note that a factual belief with 100% certainty, that a Republican will win, would have much the same effect as the antecedent of a counterfactual. If (2) were believed, not merely as very likely but with absolutely unshakable confidence, then (3) should follow, even if (1) is only believed with moderate confidence. Thus those who attributed the problem to difficulties about probability were in a way right, though this misses the point about understanding of the conditionals. This does pose a problem for classical treatments such as David Lewis's *Counterfactuals*. According to Lewis, modus ponens applied to a counterfactual conditional is valid. The argument attributed to McGee seems to refute this. P.S.: I write the above without having read Van McGee's article, on the basis of conversations and reading AIList. I intend to get to McGee's article in the near future. ------------------------------ Date: 9 Oct 85 10:55:00 EDT From: "CUGINI, JOHN" Reply-to: "CUGINI, JOHN" Subject: DO you really need an AI machine? > Date: Sun, 06 Oct 85 15:20:34 EDT > From: "Srinivasan Krishnamurthy" <1438@NJIT-EIES.MAILNET> > Subject: Do I really need a AI Machine? > > Dear readers, > I work at COMSAT LABS, Maryland. We are getting into AI in a big way > and would like comments, suggestions and answers to the > following questions,to justify a big investment on a AI machine. > > * What are the specific gains of using a AI Machine (like symbolics) > over developing AI products/packages on general purpose > machines like - VAX-11/750-UNIX(4.2BSD), 68000/UNIX etc. > ... [more questions follow] Well, I'm not qualified to give detailed answers, but let me rave mildly over some personal experience and indulge in a little heresy in hopes of provoking some discussion: So we're gonna do an expert systems project. So we got a Symbolics 3600, so that me and one other guy can develop an expert system. problem no. 1 - Geez, what a learning curve, when coming off VAX/VMS - I mean, sure, ya gotta learn expert system techniques, ya gotta learn Common Lisp, but I also gotta learn another file system and operating system (which is somehow never quite referred to as such), I gotta learn an editor with more features than I could use in a million years, I gotta learn window navigation -- after 3 months, I have 12 dense pages of cheat sheets, and am just getting to the point of adequacy (operationally defined as: not getting caught inside the inspector and being unable to figure out how to get out short of a boot-up). problem no. 2 - We've got one screen and one keyboard - even with only two people, it's surprising how often our schedules don't mesh. Cugini's tentative hypothesis of hardware-sharing: the number of people wishing to use a shared single-user system at any time is even. problem no. 3 - So now we've got an AI lab, and whenever I wanna do something I actually have to STAND UP AND WALK to the machine (which may or may not be occupied anyway) - so I find myself conversing with VMS, whose terminal sits conveniently to my right - and login takes maybe 20 seconds, as opposed to a 3-minute boot-up. problem no. 4 - and of course we don't have the Symbolics netted to the VAX yet (or ever?), and so you can kiss data-sharing good-by. Where should the (English) documentation for our expert system reside? On the VAX, where I've got editors, formatters (eg runoff), and laser printers already, or on the Symbolics, where the code lives, but which at this point has no hard-copy output? What if I get some good Lisp code over our beloved ailist? Do I key it in at the Symbolics? Well, you get the idea - there's more to doing expert systems than telling the forklift where to place your AI machine. Let me forestall some rebuttals by saying, sure I know some things we're doing wrong, and yeah, we should net the Symbolics to the VAX and yeah, we should buy a 3640 or whatever for additional users. But point 1 is there's a larger-than-I(-and-maybe-you)-suspected investment in hardware, software, and time to be able to exploit these "power tools" of AI. Point 2 is it's not so clear when you need all that performance that an AI workstation is giving you. If PC's are seen as an adequate delivery vehicle for an expert system, the assumption would seem to be that you need performance during development - but that doesn't seem right either - how fast does an editor have to be? When you're doing logical testing (as opposed to performance testing), wouldn't you be dealing with *smaller* amounts of data, than in operational mode? Why would it not make more sense to develop code on a relatively small system and then use the performance of a 3600 for large-scale logic-crunching, just as you might develop FORTRAN code on a PC, and then run it on a CYBER? I'm willing to believe I'm wrong about this, but I don't understand why. Point 3 is that the reason a lot of CompScis deride PL/I and embrace, say, Pascal, is that PL/I is big, complicated, clumsy, gives you everything you ever wanted and several that you didn't, etc etc, whereas Pascal is small, elegant, well-designed, etc etc. Any analogies here? I should say that the Symbolics itself *is* fast, reliable, and well-documented (but complicated) so I'm not complaining about Symbolics per se. This is a generic complaint. So now I'm using VAX LISP (which - shame! - does not yet have complex numbers, but is otherwise pretty good), and I find that the less-sophisticated editor is powerful enough for me, there are reasonable tools for tracing, debugging, pretty-printing, and I haven't yet been slowed down by performance problems. What am I missing here? Why am I happier on (sneer) a VAX than on a glamorous Symbolics? (Replies implying coarse sensitivity on the part of the writer will be, of course, be given the most serious consideration and then dismissed). Needless to say, these are my own utterly idiosyncratic views, and in no way reflect the policy, de jure or de facto, of the National Bureau of Standards, the Department of Commerce, or the entire Federal Government. John Cugini Institute for Computer Sciences and Technology National Bureau of Standards Bldg 225 Room A-265 Gaithersburg, MD 20899 phone: (301) 921-2431 ------------------------------ End of AIList Digest ******************** From csvpi@vpics1 Thu Oct 10 22:34:29 1985 Date: Thu, 10 Oct 85 22:34:25 edt From: csvpi@vpics1.VPI To: fox@opus (FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: R Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a011626; 10 Oct 85 13:35 EDT Date: Thu 10 Oct 1985 09:26-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #142 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Thu, 10 Oct 85 22:10 EST AIList Digest Thursday, 10 Oct 1985 Volume 3 : Issue 142 Today's Topics: Seminars - AI Meets Natural Stupidity (CSLI) & Learning Expert Knowledge (UT) & Interactive Modularity (UCB), Seminar Series - Commonsense and Nonmonotonic Reasoning (CSLI), Conference - Logic in Computer Science ---------------------------------------------------------------------- Date: Wed 9 Oct 85 16:51:08-PDT From: Emma Pease Subject: Seminar - AI Meets Natural Stupidity (CSLI) [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] CSLI ACTIVITIES FOR *THIS* THURSDAY, October 10, 1985 12 noon TINLunch Ventura Hall ``Artificial Intelligence Meets Natural Stupidity'' Conference Room by Drew McDermott Discussion led by Roland Hausser, U. of Munich McDermott discusses three `mistakes', or rather bad habits, which are frequent in A.I. work. He speaks from his own experience and cites several illuminating and amusing examples from the literature. In this TINLunch I will be discussing his thoughts on treating reference in A.I., which are discussed in the section entitled `unnatural language'. --Roland Hausser ------------------------------ Date: Tue, 8 Oct 85 16:01:56 cdt From: rajive@sally.UTEXAS.EDU (Rajive Bagrodia) Subject: Seminar - Learning Expert Knowledge (UT) Machine Learning for Acquiring Expert Knowledge by Bruce Porter noon, Friday 11th, Pai 3.38 An important effort in Artificial Intelligence is the construction of Expert Systems, but this effort is stymied by the problem of acquiring knowledge to guide problem solving and reasoning. This talk reviews efforts in Machine Learning to automate knowledge acquisition and describes our current approach to the problem. ------------------------------ Date: Wed, 9 Oct 85 16:48:25 PDT From: admin@ucbcogsci.Berkeley.EDU (Cognitive Science Program) Subject: Seminar - Interactive Modularity (UCB) BERKELEY COGNITIVE SCIENCE PROGRAM Fall 1985 Cognitive Science Seminar -- IDS 237A TIME: Tuesday, October 15, 11:00 - 12:30 PLACE: 240 Bechtel Engineering Center DISCUSSION: 12:30 - 1:30 in 200 Building T-4 SPEAKER: Ronald M. Kaplan, Xerox Palo Alto Research Center and Center for the Study of Language and Information, Stanford University TITLE: ``Interactive Modularity'' Comprehensible scientific explanations for most complex natural phenomena are modular in character. Phenomena are explained in terms of the operation of separate and indepen- dent components, with relatively minor interactions. Modular accounts of complex cognitive phenomena, such as language pro- cessing, have also been proposed, with distinctions between phonological, syntactic, semantic, and pragmatic modules, for example, and with distinctions among various rules within modules. But these modular accounts seem incompatible with the commonplace observations of substantial interactions across component boundaries: semantic and pragmatic factors, for instance, can be shown to operate even before the first couple of phonemes in an utterance have been identified. In this talk I consider several methods of reconciling modular descriptions in service of scientific explanation with the apparent interactivity of on-line behavior. Run-time methods utilize interpreters that allow on-line interleaving of operations from different modules, perhaps including addi- tional "scheduling" components for controlling the cross- module flow of information. But depending on their mathemati- cal properties, modular specifications may also be transformed by off-line, compile-time operations into new specifications that directly represent all possible cross-module interac- tions. Such compilation techniques allow for run-time elimi- nation of module boundaries and of intermediate levels of representation. I will illustrate these techniques with exam- ples involving certain classes of phonological rule systems and structural correspondences in Lexical-Functional Grammar. ------------------------------ Date: Wed 9 Oct 85 16:51:08-PDT From: Emma Pease Subject: Seminar Series - Commonsense and Nonmonotonic Reasoning (CSLI) [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] COMMON SENSE AND NON-MONOTONIC REASONING SEMINARS Organized by John McCarthy and Vladimir Lifschitz Computer Science Dept., Stanford University A series of seminars on Common Sense and Non-monotonic reasoning will explore the problem of formalizing commonsense knowledge and reasoning, with the emphasis on their non-monotonic aspects. It is important to be able to formalize reasoning about physical objects and mental attitudes, about events and actions on the basis of predicate logic, as it can be done with reasoning about numbers, figures, sets and probabilities. Such formalizations may lead to the creation of AI systems which can use logic to operate with general facts, which can deduce consequences from what they know and what they are told and determine in this way what actions should be taken. Attempts to formalize commonsense knowledge have been so far only partially successful. One major difficulty is that commonsense reasoning often appears to be non-monotonic, in the sense that getting additional information may force us to retract some of the conclusions made before. This is in sharp contrast to what happens in mathematics, where adding new axioms to a theory can only make the set of theorems bigger. Circumscription, a transformation of logical formulas proposed by John McCarthy, makes it possible to formalize non-monotonic reasoning in classical predicate logic. A circumscriptive theory involves, in addition to an axiom set, the description of a circumscription to be applied to the axioms. Our goal is to investigate how commonsense knowledge can be represented in the form of circumscriptive theories. John McCarthy will begin the seminar by discussing some of the problems that have arisen in using abnormality to formalize common sense knowledge about the effects of actions using circumscription. His paper Applications of Circumscription to Formalizing Common Sense Knowledge is available from Rutie Adler 358MJH. This paper was given in the Non-monotonic Workshop, and the present version, which is to be published in Artificial Intelligence, is not greatly different. The problems in question relate to trying to use the formalism of that paper. The seminar will replace the circumscription seminar we had last year. If you were on the mailing list for that seminar then you will be automatically included in the new mailing list. If you would like to be added to the mailing list (or removed from it) send a message to Vladimir Lifschitz (VAL@SAIL). The first meeting is in 252MJH on Wednesday, October 30, at 2pm. ------------------------------ Date: Wed 9 Oct 85 16:51:08-PDT From: Emma Pease Subject: LICS Conference [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] LICS CONFERENCE A new conference, LICS, (an acronym for ``Logic in Computer Science'') will meet in Cambridge, Mass, June 16-18, 1986. The topics to be covered include abstract data types, computer theorem proving and verification, concurrency, constructive proofs as programs, data base theory, foundations of logic programming, logic-based programming languages, logics of programs, knowledge and belief, semantics of programs, software specifications, type theory, etc. For a local copy of the full call for papers, contact Jon Barwise (Barwise@CSLI) or Joseph Goguen (Goguen@SRI-AI), members of the LICS Organizing Committee. ------------------------------ End of AIList Digest ******************** From csvpi@vpics1 Mon Oct 14 05:33:59 1985 Date: Mon, 14 Oct 85 05:33:55 edt From: csvpi@vpics1.VPI To: fox@opus (FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: R Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a000736; 14 Oct 85 0:47 EDT Date: Sun 13 Oct 1985 20:43-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #143 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Mon, 14 Oct 85 05:17 EST AIList Digest Monday, 14 Oct 1985 Volume 3 : Issue 143 Today's Topics: Queries - Autonomous Vehicles & YAPS & Prolog vs. OPS5 & ES Tools & Franz Lisp Behavior, Bindings - Symbolics Lisp Machine Mailing List, Corrections - Verification Peer Review & TIMM Expert System, AI Tools - Workstations ---------------------------------------------------------------------- From: C. Ian Connolly Subject: Autonomous Vehicles Apropos a recent AILIST entry: Does anyone have more information on the DARPA Autonomous Vehicle demo that Waxman, et al (I think - correct me if I'm wrong) gave earlier this year? I'd *love* a review, if anyone was there that can send one out on this list. What speeds were it capable of, what methods were used, etc...? ------------------------------ Date: Wed, 09 Oct 85 18:30:49 EDT From: "Srinivasan Krishnamurthy" <1438@NJIT-EIES.MAILNET> Subject: YAPS - Commercial Version Info. Dear Folks! Does anybody have information on the Commericial Availability of a Production System called "YAPS"? This was developed at the University of Maryland and funded by the Goddard Space Flight Center. I can't seem to find the right people to talk to at Maryland, regarding this system. Any leads will be greatly appreciated... Thanks in Advance. .....Vasu ------------------------------ Date: 10 Oct 85 10:02:58 PDT (Thursday) From: Cornish.EIS@Xerox.ARPA Subject: Prolog vs. OPS5 Can anyone provide me with a compare-and-constrast discussion of Prolog vs. OPS5. To use an analogy from this list, are they both screwdrivers or both chisels ? Jan ------------------------------ Date: Thu 10 Oct 85 08:53:24-PDT From: Mark Richer Subject: es tools query My list of major commercial AI software tools includes: (1) S.1., (2) KEE, (3) ART, and (4) Knowledge Craft There is also DUCK which seems to be more like an enhanced logic programming language than the kind of tools (1)-(4) represent. There is TIMM which was blasted on this list recently. One other candidate appears to be KES or KES2. Do you have any comments on this system? Is it a strong competitor to the other major tools? mark ------------------------------ Date: 10 Oct 1985 14:37-CST From: leff%smu.csnet@CSNET-RELAY.ARPA Subject: franz lisp ? The following demonstrates that enabling the trace facility causes return values from lisp functions consisting of prog bodies to be set to nil. Is this supposed to happen and what does one do about it? % cat blah1.l (defun Blah (x y z a) (prog (v) (setq y (add 3 5)) (return 1) )) % lisp Franz Lisp, Opus 38.79 -> (load "blah1.l") [load blah1.l] t -> (setq A (Blah 3 5 7 9)) 1 -> (step e) [autoload /usr/lib/lisp/step] [load /usr/lib/lisp/step.l] t -> (setq B (Blah 3 5 7 9)) (setq B (Blah 3 5 7 9)) (Blah 3 5 7 9) 3 5 7 9 (prog (v) (setq y (add 3 5)) (return 1)) (setq y (add 3 5)) (add 3 5) 3 5 8 8 (return 1) 1 nil nil nil -> ------------------------------ Date: Fri 11 Oct 85 12:43:55-PDT From: Richard Acuff Subject: Re: Symbolics lisp machine mailing list ?? [Forwarded from the Stanford bboard by Laws@SRI-AI.] The Symbolics Lisp Users' Group has a mailing list SLUG@R20.UTexas.ARPA. This can be read from the BBoard SYMBOLICS-LISP-USERS on Sumex. Send mail to SLUG-Request@R20.UTexas.ARPA to get on the list. There is also BUG-LISPM@MIT-MC which may be of some interest (also a BBoard on Sumex), as well as {Bug,Info}-TI-Explorer@Sumex for TI Explorers. Again, use the -Request convention for getting added. -- Rich ------------------------------ Date: Wed, 9 Oct 85 21:38:01 PDT From: Dick Kemmerer Subject: Peer Review A colleague forwarded a copy of AIList Digest #133 to me. In this issue Michael Melliar-Smith's response to John Nagle's SIFT verification message contains a reference to a verification study that is being sponsored by the DoD Computer Security Center. I am the PI for this study and would like to comment that the reference is somewhat misleading. In particular, the study group did not look at the SIFT work. Also, the suite of tools that we reviewed were the enhanced HDM tools (most of which have been developed since the SIFT work). ------------------------------ Date: Thu, 10 Oct 85 12:22:01 edt From: decvax!linus!raybed2!gxm@UCB-VAX.Berkeley.EDU (GERARD MAYER) Subject: TIMM Correction I'd like to point out the only error I found in Touretzky's article about TIMM. According to the General Research technical and sales people at the IJCAI-85 booth TIMM CANNOT call any external functions: Fortran or otherwise. I found this amazing and asked several of their people the same question. I have sent a synopsis of my review of TIMM, KEE, ART, etc. to Cowan directly. Gerard Mayer Raytheon Research Division uucp ..linus!raybed2!gxm ------------------------------ Date: 10 Oct 85 11:01:17 PDT (Thu) From: Jeff Peck Subject: RE: do you REALLY need a AI machine? Funny how when you say a "lisp machine" some people just assume you mean "Symbolics". My experience has been that you can avoid most of the problems John Cugini complained about by chosing an alternate vendor. For instance, with a $100K investment (equivalent to your first 3600) you can buy 3 or 4 Xerox lisp machines. So, now you can put one in each programmers office (and not have to walk down the hall to an occupied machine), no standing in line. The Xerox machines network to VAX/VMS or Unix file servers, so sharing is easy. The Xerox user interface is so transparent that you can learn it in a day or two (the editor in about 5 minutes). And, although the Xerox may be 1/3 the speed of a 3600, as John points out, "who cares" when you are just building and exploring? If you get into serious production of large systems, then move it to a Symbolics, or a Cray. If you are doing Research on AI technology, you may also want the faster system, but most industrial labs seem to be more into applications development. (Also, the Xerox machines support (or soon will) both Quintus Prolog, and CommonLisp, integrated with the InterLisp-D environment). As for a VAX it's never fast enough. If all you do is editing, maybe; but where are the graphics? And timesharing LISP is a serious development problem: "Is my program stuck, or is someone else just compiling something?" For $250K, you can do a lot better with personal workstations. This is not intended to be a plug for Xerox, and of course, these are just my personal observations. j. peck ------------------------------ Date: Wed, 9 Oct 85 20:56:49 PDT From: Richard K. Jennings Subject: Big Lisp Machines We, as are you, are involved in applying AI to our jobs. We have 5 PC-AT's and a Symbolics (which we got from another organization). The PC's are much more useful, and a VAX would be ideal. The basic reason is that PC-AT's and VAXEN are not that much slower than the big machines, and cost less, and are compatible with alot more applications programs and peripherals. Most applications need a big data base to work of, and the AI portion is relatively small. If you already have a great deal of LISP code already written, and you already have VAXEN and ethernet, then a big machine might be OK. My advice is to purchase AT's and micro-vax II's with your bucks, and I sincerely doubt you will regret it. Jed Marti (ARPA: marti@rand-unix.arpa ATT:213 393-0411) published a benchmark of many machines (including a Symbolics 3600, VAX 11/780, 750) running REDUCE problems (a symbolics math package). I think the 780 was a tad faster than the 3600 and the 750 a tad slower (~10%). If you are working on the part of COMSAT which flies satellites, I would like to find out what you are doing -- that's what we are using AI (planning to use AI) for. Hope these comments help, Richard Jennings, Air Force Satellite Control Facility Sunnyvale CA 95051 ATT: 408 744-6427 ARPA: jennings@aerospace ------------------------------ Date: Thu, 10 Oct 85 16:38 EST From: "Christopher A. Welty" Subject: AI Workstations I don't intend to offend anyone, but after reading John Cugini's response to Srinivasan Krishnamurthy's query about AI machines I was a little taken aback. The reason one gets an AI machine is because it is the most powerful AI/ES development tool around. It provides the developer with facilities that make the entire system development process a hundred times easier (this number may vary depending on what system you are coming from). Of course you have to learn how to use it. Some people look forward to the chance to learn something new, others prefer to know only one or two systems and hail them as the ultimate. Those who prefer to learn know that each different system has a use that makes it helpful for certain applications, and a hinderance in others. If you really make an effort to learn how to use an AI workstation, you will find (especially if you've had to do development on other systems) that you will be far more productive, and you will be doing things more in the way they should be done. We all know that it is often easier to cheat than to do things the right way, and often times cheating makes later development more difficult. With the extensive support environments for AI/ES that these workstations provide, doing things the right way is made easier (almost easier than cheating). From Mr. Cugini's statement, his objections to the workstations seemed no more than laziness...and that seems no reason to dissuade others from getting them. If you are really getting into AI "in a really big way," an AI workstation is a must. You won't know what your missing if you never get one, and if you do (and take the time to learn it) you won't know how you did without it. -Christopher A. Welty RPI/CIE Systems Mgr ------------------------------ End of AIList Digest ******************** From comsat@vpics1 Tue Oct 15 00:01:05 1985 Date: Tue, 15 Oct 85 00:01:01 edt From: comsat@vpics1.VPI To: fox@opus (MILLER,FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: RO Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a001216; 14 Oct 85 2:10 EDT Date: Sun 13 Oct 1985 20:58-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #144 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Mon, 14 Oct 85 23:05 EST AIList Digest Monday, 14 Oct 1985 Volume 3 : Issue 144 Today's Topics: News - Grace Murray Hopper Award, Intelligence & Learning - An Appreciation of Our Own Make-up, Archive Services - BITNIC Server for Recent Issues & VPI Full Archive (Micro LISP Search) ---------------------------------------------------------------------- Date: Thu 10 Oct 85 22:49:02-PDT From: Ken Laws Subject: Grace Murray Hopper Award >From the October issue of CACM: Cordell Green of Kestrel Institute was chosen "for establishing a theoretical basis for logic programming and providing a resolution theorem proven to carry out a programming task by constructing the result which the computer is to compute. For proving a constructive technique correct and for presenting an effective method for constructing the answer; these contributions providing an early theoretical foundation for Prolog and logic." ------------------------------ Date: Fri, 11 Oct 85 11:42:41 GMT From: gcj%qmc-ori.uucp@ucl-cs.arpa Subject: ``An Appreciation of Our Own Make-up'' >From the Guardian newspaper, 10 October 1985:- ``What he (Alan Kay) would like to see AI people trying to build is "not superhumans or humans, but mammals," contrivances that can explore and learn but do not have to use language or learn differential calculus." Ultimately, he asserts, "the basic end of AI research is an appreciation of our own make-up." '' I find the idea that these "contrivances" do not have the need to make use of language rather strange, since there must be some form of communication required between us and them and that will be termed language. The more important message of the short quote is the repetition of the idea that intelligence is learning, starting from square one and building a model of the world that is in one's view, and the view from the chip must be through the language of data. Gordon Joly gcj%qmc-ori@ucl-cs.arpa ------------------------------ Date: Thu, 10 Oct 85 12:01 EDT From: Henry Nussbacher Subject: Database service available for back issues of Ai-List This is to announce that some new services have been added to the inter-network server running at Bitnic. Certain selected Arpanet digests are now being loaded into a Spires database and are therefore searchable from anywhere as long as you can send RFC822 mail. If you are interested in using this service, send a piece of mail to: DATABASE%BITNIC.BITNET@WISCVM.ARPA or DATABASE%BITNIC.BITNET@WISCVM.WISC.EDU and have as the first 3 lines of your file (case does not matter): help help arpanet help design The server will send back to you 3 help files describing how to use the internet server, how to search Arpanet digests and how the whole thing was designed. Read over the section on "Signup" carefully before making further use of the Database server. Presently, the following 5 Arpanet forums are being loaded into the Database: Name Retention Period ============= ================ Ai-List 2 months Info-Ibmpc 2 months Info-Mac 2 months Info-Graphics 3 months Info-Nets 3 months The retention period is set for a short duration in order to see if Bitnic can handle the volume of data that needs to be stored in Spires. This service was initialized on October 4th, 1985 so currently there are just a few items available in the Database. Example of search command: FIND TEXT UNIX (IN INFO-IBMPC TABLE would find all entries in Info-Ibmpc that contain the word UNIX. An entry is just the section within a "digested" digest that makes reference to the word UNIX. For further details read over the help files. Henry Nussbacher (Hank@Bitnic.Bitnet) Bitnet Development and Operations Center ------------------------------ Date: Wed, 9 Oct 85 18:39 EST From: Ed Fox Subject: reply to query on micros and LISP From: france (Robert France) Date: Mon, 7 Oct 85 14:35:52 edt To: fox, sharan Subject: Re: chance to do a useful search Ken: Robert Blum on 3 Oct asked for information on "currently marketed LISPs for micros" including pointers to review articles. While our new system that will classify and work with components of messages is not yet ready, our adaptation of the SMART system is running and ready for queries just like this. Robert France did a search with the following results. Feel free to publish in AIList or to send directly to Blum. Send other queries along too! I need then for experimentation, and only request that the author of the query be willing to tell me which messages are relevant to the question. Thanks, Ed Fox [At the risk of having this message show up in all future searches, I've decided to pass it along. I deleted one false hit (a query from Rene Bach), and I remember at least one other very lengthy Lisp review that was not found. -- KIL] _______ .I 418 .W Tuesday, 20 Sep 1983 .V Volume 1 .U Issue 59 .D Mon, 19 Sep 1983 11:41 EDT .N .A WELD%MIT-OZ@MIT-MC .S Micro LISPs .B For a survey of micro LISPs see the August and Sept issues of Microsystems magazine. The Aug issue reviews muLISP, Supersoft LISP and The Stiff Upper Lisp. I believe that the Sept issue will continue the survey with some more reviews. Dan _______ .I 1113 .W Thursday, 8 Mar 1984 .V Volume 2 .U Issue 27 .D Tue 6 Mar 84 15:48:55-PST .N Sam Hahn .A SHahn@SUMEX-AIM.ARPA .S IQLISP Source .B The source for IQLisp is: Integral Quality, Inc. P.O. Box 31970 Seattle, WA 98103 (206) 527-2918 Claims to be similar to UCI Lisp, except function def's are stored in cells within identifiers, not on property lists; arg. handling is specified in the syntax of the expression defining the function, I/O functions take an explicit file argument, which defaults to the console; doesn't support FUNARGS. IQLisp does provide: 32kb character strings, 77000 digit long integers, IEEE format floating point, point and line graphics, ifc to assembly coded functions, 31 dimensions to arrays, Costs $175 for program and manual, PCDOS only. I've taken the liberty to include some of their sales info for those who may not have heard of IQLisp. It's fairly new, and they claim to soon make a generic MSDOS version (though probably without graphics support). _______ .I 1264 .W Thursday, 12 Apr 1984 .V Volume 2 .U Issue 45 .D 11 Apr 1984 0206 PST .N Reply-to: LARRY@JPL-VLSI.ARPA .A Larry Carroll A.C. McIntosh, USADACS@STL-HOST1. _______ .I 2512 .W Sunday, 20 Jan 1985 .V Volume 3 .U Issue 5 .D Thu 17 Jan 85 00:33:35-PST .N Sam Hahn .A SHahn@SUMEX-AIM.ARPA .S Lisp for PC .B If you're using PC's and looking for a Lisp, I'd suggest TLC-Lisp, from The Lisp Company. I myself have not used GCLisp, but have been quite impressed with TLC-Lisp, which has a compiler, an object-class system, packages, auto-load entities, and costs less than half what GCLisp costs. TLC is John Allen's (The Anatomy of Lisp) company, located in Redwood Estates, CA. I have no connection with TLC except as a customer. -- sam hahn _______ .I 2773 .W Friday, 8 Mar 1985 .V Volume 3 .U Issue 31 .D Thu 7 Mar 85 08:44:14-PST .N Ken Laws .A Laws@SRI-AI.ARPA .S The Artificial Intelligence Report .B Ted Markowitz recently asked about newsletters. [...] The following are the topics covered in back issues of The Artificial Intelligence Report. I'm told that back issues are still available, but I don't know the price. [...] Vol. 1, No. 3, March, 1984 AI and the Personal Computer: Expert systems, natural language, LISP; [...] Vol. 2, No. 3, March, 1985 LISP on the PC: TLC LISP, GCLISP; [...] [...] This newsletter was the first one mentioned in AIList. Since that time, it has moved from Los Altos to: Artificial Intelligence Publications Suite Three 3600 West Bayshore Road Palo Alto, CA 94303 - 4229 U. S. A. (415) 424-1447 -- Ken Laws _______ You might try using other keywords (names of micros?) or going further. -- Robert ------------------------------ End of AIList Digest ******************** From csvpi@vpics1 Mon Oct 14 05:33:52 1985 Date: Mon, 14 Oct 85 05:33:48 edt From: csvpi@vpics1.VPI To: fox@opus (FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: R Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a001506; 14 Oct 85 3:42 EDT Date: Sun 13 Oct 1985 21:08-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #145 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Mon, 14 Oct 85 05:19 EST AIList Digest Monday, 14 Oct 1985 Volume 3 : Issue 145 Today's Topics: Opinion - Military Support & AI Hype & CS and AI Definitions, AI Tools - Workstations & Lisp vs Prolog Implementation Facts, Call for Papers - IEEE Systems, Man and Cybernetics ---------------------------------------------------------------------- Date: Thu 10 Oct 85 12:07:50-PDT From: Rich Alderson Subject: re: military flame Like it or not, the military CAN take credit for the interstate highway system, and the "civilian highway system in place before the interstates" as well--at least, I ASSUME that you are referring to the U. S. highway system. Both were built under the mandate of the Constitution of the United States (Philadelphia, 1787), which did not view "internal defence [as] a standard rationale for better internal transportation" but rather saw good internal transportation as vital to both the internal and external defense of a nation. Please note that the above is not a matter of opinion. My own opinions are just that, and I've had enough grief in my life for expressing them publicly. Obviously, I make NO claim to represent anyone else's opinions if I refuse to state my own. Rich Alderson ------------------------------ Date: Thu 10 Oct 85 10:12:48-PDT From: WYLAND@SRI-KL.ARPA Subject: AI Hype is unavoidable Part of the problem of AI hype is unavoidable: it is a result of the definition of the field, which assumes that, once the problem is defined, its solution is trivial. Artificial Intelligence and Computer Science represent two complementary approaches to the use of computers. AI is problem oriented, and CS is solution oriented. In CS, one assumes that the problem is (or is capable of being) well understood, and the task is to design a good (clean, efficient, fast, etc.) solution to the problem. In AI, the underlying assumption is that the solution will be trivial once the problem is understood, and that the task is to understand the problem. These two approaches are reflected in the names of the field. Computer Science is the science of using computers, a clear, hard, objective definition about the use of a tool to solve problems. Artificial Intelligence is concerned with the simulation of intelligence, a fuzzy, open, subjective definition about the study of the problem of simulating a human behavior called intelligence for which there is no clear, generally accepted definition. Each field encounters problems when its assumption is violated. In CS, programming disasters can result if you start coding before you have defined the problem. Therefore, CS has developed structured programming, programming specifications, etc. tools to insure that the problem *is* well defined before the solution is attempted. In AI, trouble starts when, after the problem is understood, the solution is *not* trivial in terms of computer performance, such as execution time and memory space. Then, you start hacking the algorithms in order to get results in a finite amount of time and memory, hoping that you are not betraying your understanding of the problem. Therefore, AI has developed extensive editors, debuggers, windows, etc. in an attempt to insure that the implementation of the solution remains trivial. I believe that much of the AI hype problem stems from the unstated assumption of trivial solutions. Profound and impressive insights expressed as toy problems typically do not "scale up" to the real world. An AI researcher involved in this situation is embarrassed but not humiliated: the original research on the problem is still valid; there is just a "temproary problem in creating a practical solution." This obviously creates frustration for the customers who thought they were buying a solution of the problem rather than an understanding of it. If the above is true, the problem of AI hype will not go away until the field develops enough solid understanding of the problem of intelligence to change its name from AI to a solution oriented name, like Machine Learning, etc. We are probably in the same position now that physical science was before Galileo and Newton when it was called Natural Philosophy - full of metaphysics, passionate argument, and conflicting data (i.e., where the action is). Dave Wyland ------------------------------ From: CONNOLLY CHRISTOPHER IAN Subject: AI Definition and Tools 1) I can't help but think that a cause of the recent arguments on AI hype rests in the question "What is AI?". Note that I say >>**"A"**<< cause. Definitions, anyone? 2) AI Machines - My observation is that the startup time on a 3600 without help is quite long. There are a few people here who have taken a Symbolics Lisp course and seem to have picked up on the stuff much more quickly (2 or 3 months?). Once you know how to use the machine though, I think it's a far better programming environment than VAX/VMS. I've seen nothing on a VAX that parallels the Window Debugger (wherein the entire stack can be dissected), the Inspector (wherein your data structures can be dissected), and the Flavor Examiner (wherein your data types can be inspected). The latter two are also a great help when you have no source. I think it speeds up my programming by a factor of 5, at least. Anyway, that's yet another opinion... ------------------------------ Date: Sat, 12 Oct 85 12:28:27 EDT From: George J. Carrette Subject: Lisp vs prolog, implementation facts Actually, all three of the Symbolics, LMI, and TI lispmachines give the lisp-level system programmer access to extremely low level data type and stack operations, there is no need to go to microcode for that sort of thing. The LM-PROLOG that I maintain from time to time in the Symbolics and LMI environment creates its own datatypes by the usual lisp punning techniques, fooling with locatives, forwarding pointers and the internals of CDR-CODED lists without needing microcode support, and amazingly keeping a good deal of transportability. Such is the ubiquity of certain hacks of lispmachine implementation. Microcode is used, optionally, only for hand-coding functions that are also written in lisp. The LM-PROLOG technique is in the class of CONTINUATION PASSING techniques of prolog implementations, as described for example in a chapter of Sussman and Ablesons "Structure and Interpretation ..." book. This kind of technique has more overhead associated with creation of real function evaluation frames and such, (at least on a simple stack-machine) and observably gets about 1/2 or 1/3 the BENCHMARK performance of a lower-level machine-model implementation such as described by Warren. Overall performance of a practical prolog "expert-system" may depend more on virtual memory performance considerations of course rather than what happens in the register-usage-mostly situation of some benchmarks. Also consider that commercial systems put into production often have assembly language coding of important subroutines, so high-level/low-level language interface issues are important. The continuation passing technique provides a more natural and efficient interface to "assembly language" (i.e. LISP on a LISPMACHINE) than other models. When talking about a commercial lispmachine it is important to think in terms of LISP as a COMPUTER ENGINEERING technique rather than as having anything to do with AI programming in particular. ------------------------------ Date: 12 Oct 1985 11:04-EDT From: milne@wpafb-afita Subject: call for papers - IEEE SMC CALL FOR PAPERS IEEE TRANSACTIONS ON SYSTEMS, MAN AND CYBERNETICS. SPECIAL ISSUE ON Causal and Strategic Aspects of Diagnostic Reasoning Papers are solicited for a special issue of IEEE Transactions on Systems, Man and Cybernetics that will be devoted to the topic, "Causal and Strategic Aspects of Diagnostic Reasoning." Dr. Robert Milne, Army Artificial Intelligence Center will be the guest editor of the special issue. While it is expected that the research to be reported will be typically backed up by concrete analyses or system building for real-world diagnostic problems, the intent is to collect the most sophisticated ideas for diagnostic reasoning viewed as a generic collection of strategies. Articles should attempt to describe the strategies in a domain-independent manner as much as possible. Articles that merely describe a successful diagnostic expert system in a domain by using well-known languages or strategies will typically not be appropriate. Papers reporting on psychological studies, epistemic analyses of the diagostic process, elucidating the strategies of first-generation expert systems, descriptions of specific diagnostic systems that incorporate new ideas for diagnostic reasoning, learning systems for diagnosis are some examples that will be appropriate. It is expected that most articles will typically concentrate on some version or part of the diagnostic problem, so it is important that the paper state clearly the problem that is being solved independent of the implementation approaches adopted. Papers will be reviewed carefully by referees selected by the Transactions Editorial Board. Five copies of the manuscript should be submitted to Dr. Robert Milne at the following address by January 15. It will be helpful if people who intend to submit manuscripts for consideration can let the guest editor know of their intent as soon as possible via arpanet or telephone. Submit papers by January 15th, 1986 to: Dr. Robert Milne US Army AI Center HQDA DAIM-DO Washington, D.C. 20310-0700 phone:(202)-694-6913 arpa: milne@wpafb-afita Author's Timeline: 15 January 1986 Papers Due 15 April 1986 Notificationo of acceptance/rejection 1 June 1986 Final Manuscripts due November 1986 Publication ------------------------------ End of AIList Digest ******************** From comsat@vpics1 Tue Oct 15 22:09:47 1985 Date: Tue, 15 Oct 85 22:09:42 edt From: comsat@vpics1.VPI To: fox@opus (MILLER,FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: RO Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a009005; 15 Oct 85 1:05 EDT Date: Mon 14 Oct 1985 21:20-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #146 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Tue, 15 Oct 85 21:59 EST AIList Digest Tuesday, 15 Oct 1985 Volume 3 : Issue 146 Today's Topics: AI Tools - YAPS Info, Bindings - Scheme Mailing List, Opinion - Scaling Up, Psychology & Logic - Modus Ponens ---------------------------------------------------------------------- Date: 14 Oct 85 15:58:27 EDT (Mon) From: Liz Allen Subject: YAPS info In response to the question about obtaining YAPS: YAPS is available from the Univ of Maryland along with some other packages like a flavors package. They run under Franz Lisp (we supply a slightly hacked version of Opus 38.91) and on Vaxes running Berkeley Unix. For more information, send mail to me. -Liz liz@tove.umd.edu or liz@maryland.arpa ------------------------------ Date: Mon, 14 Oct 85 20:37:24 EDT From: Hal Abelson Subject: Scheme mailing list Scheme@MIT-MC.ARPA is a network-wide mailing list for discussions concerning the Scheme dialect of Lisp -- both as a vehicle for investigating language development and as a vehicle for teaching about computer science. To be added to the list, please send mail to Scheme-Request@MIT-MC.ARPA. Remote sites with many entries are encouraged to set up local distribution lists.. ------------------------------ Date: Mon 14 Oct 85 07:50:00-PDT From: Gary Martins Subject: Scaling Up In a recent issue of AIList [#145], Dave Wyland expresses a rather common epistemological error, in his attempt to defend "AI" as we know and love it today -- hype and all ! Mr. Wyland seems to think that finding problem solutions which "scale up" is a matter of manufacturing convenience, or something like that. What he seems to overlook is that the property of scaling up (to realistic performance and behavior) is normally OUR ONLY GUARANTEE THAT THE "SOLUTION" DOES IN FACT EMBODY A CORRECT SET OF PRINCIPLES. To put it more simply, if the solution doesn't scale up, then it just plain isn't a solution, even if the inventor feels he has made a lot of "profound and impressive insights". The point is, these insights aren't very profound or impressive (except perhaps to IJCAI referees) if they fail the scaling test. This principle is generally understood by persons with real engineering backgrounds, but seems to come as news to "AI" folks. G.R. Martins ------------------------------ Date: Wed, 9 Oct 85 16:43 EST From: Mukhop Subject: Non-contradictory set of beliefs and dependencies > Date: Thu, 3 Oct 85 18:28:03 PDT > From: albert@UCB-VAX.Berkeley.EDU (Anthony Albert) > Subject: Re: Counterexample to modus ponens > > As far as beliefs, a non-contradictory set could be: > > 1) If a Republican wins then Reagan will win. > 2) If Reagan doesn't win then a Republican won't win. > 3) Unless a Democrat wins, a Republican will win. > Beliefs 1) and 2) of this non-contradictory set are one and the same as far as inferential power is concerned: P => Q <=> ~Q => ~P Also, this set of non-contradictory beliefs does not include the notion that Reagan (or a Republican) will probably win. It merely states that Reagan's chances are better than his Republican opponent. Getting back to the original set of statements: > (1) If a Republican wins the election then if the winner is not Ronald > Reagan, then the winner will be John Anderson. > > (2) A Republican will win the election. > > Yet few if any of these people believed the conclusion: > > (3) If the winner is not Reagan then the winner will be Anderson. > My perception of the commonly held beliefs at that time is: a) There are three contestants in the election. b) Ronald Reagan has the highest probability of winning. c) John Anderson has the lowest probability of winning. d) John Anderson and Ronald Reagan are the only Republicans contesting the elections. The second statement in the original set ("A Republican will win the election") was a commonly held belief, arrived at from: - Ronald Reagan has the highest probability of winning, and - Ronald Reagan is a Republican. (Of course, John Anderson increased the odds in favor of a Republican) If the assumption is now made that Ronald Reagan will not win the election, then one can no longer make the assumption that a Republican will win. The conclusion, " If the winner is not Reagan then the winner will be Anderson," can no longer be made. It seems that the dependencies of the belief structures need to be taken into account in order to avoid contradictions. If it was reasonable to believe that a Republican would win, irrespective of the chances of Ronald Reagan, then it would be reasonable to believe: "If the winner is not Reagan then the winner will be Anderson." ------------------------------ Date: Sat, 12 Oct 1985 21:09 EDT From: MINSKY%MIT-OZ@MIT-MC.ARPA Subject: Republicans and Probability About Republicans and probability. That paradox comes from all those causes -- ambiguities, shifts in meanings from "intension" to "extension," and so forth. In my view, adult psychology is too complicated to treat in terms of such simple mathematical models as deduction and probability. You've heard my complaints about logic too often to repeat, and surely everyone has read the critiques of Kahnemann and Tversky about applications of probabilistic models to human beliefs and reasoning. I suggest another, simple example to examine. If you tell a "typical" person that "Most A's are B's" and that "Most B's are C's," then the most common inference is that "Most A's are therefore C's." More sophisticated people may say --"No, not most, but at least 26 percent of A's are C's, because they'll notice that "most" might mean 51 percent. Very few people will recognize that it is possible that no A's are C's, or be able to construct a counterexample. ------------------------------ Date: Sun, 13 Oct 85 13:39:09 edt From: pugh@GVAX.CS.CORNELL.EDU (William Pugh) Subject: Bayesian inference or nested assumptions Continuing on the subject of the Modus Ponens example, I have worked out some results on using Bayesian Inference for nested assumptions: Notation: A|B means assuming B is true, A is true if B is false, the statement is neither true nor false, it is untested. P(X) - the probability of X e.g. P(A|B) = the probability of A being true, given that B is true For the original example: >> (1) If a Republican wins the election then if the winner is not >> Ronald Reagan, then the winner will be John Anderson. >> >> (2) A Republican will win the election. >> >> Yet few if any of these people believed the conclusion: >> >> (3) If the winner is not Reagan then the winner will be Anderson. >> Let RW stand for "A Republican wins" Let RR stand for "Ronald Reagan wins" Let JA stand for "John Anderson wins" We have: 1) P((JA|~RR)|RW) = 1 2) P(RW) = very high 3) P(JA|~RR) = ?? Well, nested assumptions don't really work with the normal Bayesian calculations, so we first have to convert (1) to a normal form. To convert to a normal form, P((A|B)|C) = P(A|(B&C)) You can show this formally, but informally in English: Assuming that C, then assuming that B, then A is the same as Assuming that C and B, then A Alright, so now we have P(JA|(~RR&RW)) = 1. What can we do with this?? P(JA)P((~RR&RW)|JA) P(JA|(~RR&RW)) = ------------------- P(~RR&RW) P(JA)P(RW|JA)P(~RR|(RW&JA)) = --------------------------- P(RW)P(~RR|RW) P(JA)P(RW|JA)P(~RR|(RW&JA)) so P(JA|(~RR&RW))P(RW) = --------------------------- P(~RR|RW) P(JA) (since there can be only one winner) = --------- P(~RR|RW) Which, by other real world knowledge, you can reduce to P(RW). Side note: I was explaining this problem to a friend who does not have a background in logic. I when I told her that A => B is true when A is false, she said "That's stupid... No wonder logicians have trouble with the real world." :-) One moral of this story: Be careful of "if" in english - it oftens means something other than the standard logical meaning. Bill Pugh Cornell University ..{uw-beaver|vax135}!cornell!pugh 607-257-6994 ------------------------------ Date: Sun, 13 Oct 85 20:02:19 -0200 From: Eyal mozes Subject: Re: Counter-example to Modus Ponens >> Before the 1980 presidential election, many held the two beliefs >> below: >> (1) If a Republican wins the election then if the winner is not >> Ronald Reagan, then the winner will be John Anderson. >> (2) A Republican will win the election. >> Yet few if any of these people believed the conclusion: >> (3) If the winner is not Reagan then the winner will be Anderson. > I would say the problem with this analysis is that people believed, > instead of statement 1, the following similar statement: > (1a) If a Republican wins the election and the winner is not > Ronald Reagan, then the winner will be John Anderson. In classical philosophical logic, a conditional proposition (i.e., a proposition of the form 'if A then B') asserts the necessity of a certain sequence between two statements; the truth of 'if A then B' does not depend on the truth of A and of B, but on the connection between them - whether B's truth FOLLOWS NECESSARILY FROM A's truth. For example, the statement 'if you are alive, then you are now reading an ARPANET message' is FALSE; the condition and the consequent are both true, but the consequent does not follow necessarily from the condition (you could be alive and still be doing something else now). Now, let us look at the meaning of (1a), (2) and (3). (1a) asserts that if the winner is a Republican but not Reagan, it follows necessarily that it will be Anderson. Given that there were only two Republican candidates, (1a) is true, and everyone knew it is true. The actual results of the elections determined the truth of the two components (both turned out to be false), but made no difference about the necessary connection between them. (2) is a simple categorical proposition; before the elections, many people believed it is true, others believed it is false; it turned out to be true. (3) asserts that if the winner is not Reagan, it FOLLOWS NECESSARILY that it will be Anderson. This is obviously false, and I doubt if anyone ever believed it. Again, the actual results of the elections determined the truth of the two components (both turned out to be false), but made no difference about the lack of any necessary connection between them. Given the classical logical interpretation, then, (3) does not follow from (1a) and (2). People who believed (1a) and (2) but not (3) were perfectly consistent (and they were also correct). This example demonstrates one of the serious weaknesses of predicate calculus. Predicate calculus has no way to express necessary connections (mainly because its originators, Russell and Whitehead, held a philosophy which denies the existence of such connections); the result is the truth-functional interpretation of conditional propositions, which leads to such anti-common-sense results as making (3) follow from (1a) and (2). As for (1), I'm not sure about its meaning, but it certainly doesn't mean exactly the same as (1a). As far as I know, in all discussions, by classical logicians, of conditional propositions or of Modus Ponens, they only dealt with the case in which both components of the conditional are simple categorical propositions. It is an interesting question whether Modus Ponens remains valid in other cases as well (and this, of course, depends on what exactly such 'multiple-conditional' propositions mean). Eyal Mozes BITNET: eyal@wisdom CSNET and ARPA: eyal%wisdom.bitnet@wiscvm.ARPA UUCP: ..!decvax!humus!wisdom!eyal ------------------------------ End of AIList Digest ******************** From csvpi@vpics1 Tue Oct 15 06:48:42 1985 Date: Tue, 15 Oct 85 06:48:39 edt From: csvpi@vpics1.VPI To: fox@opus (MILLER,FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: RO Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a009480; 15 Oct 85 2:34 EDT Date: Mon 14 Oct 1985 21:28-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #147 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Tue, 15 Oct 85 06:35 EST AIList Digest Tuesday, 15 Oct 1985 Volume 3 : Issue 147 Today's Topics: Reviews - Canadian Artificial Intelligence 4 and 5 & Spatial Data Handling and Graphics Interface Conferences, Seminars - Intelligent Electronic Mail (MIT) & Connectionist Learning (GTE) & Connectionist Learning (SU) & Probabilistic Interpretation of Certainty Factors (SU) ---------------------------------------------------------------------- Date: 12 Oct 1985 10:59-CST From: leff%smu.csnet@CSNET-RELAY.ARPA Subject: Canadian Artificial Intelligence 4 Summary of Canadian Artificial Intelligence 4, June 1985 Should CSCSI/SCEIO Attempt to Influence National Policy (discusses whether their organization should work to influence national policy on AI within the Canadian government.) AI Research Spending and politics Discusses Canadian research spending and talent shortage. The Canadian AI group has been active in opposition to Star Wars. Discussion of an episode of Magnum P. I. which featured an AI researcher who developed some formula that would tip the balance in favor of whoever had it. The formula was "3 bracket prompt semicolon." The Canadian National Research Council is starting an inventory of Canadian robotics research Announcement of Babbage and Lovelace's BASIC package to expand other BASIC programs to do natural language parsing. Runs on an IBM PC with 64 to 96 K. French Article on AI and Cognitive Science at the University of Montreal Canadian Companies Applied AI Systems, Kanata Ontario doing consulting, marketing of software, custom software. Review of LOGICware which sells MPROLOG. Book reviews of "The Comercial Application of Expert System Technology" "Artificial Intelligence: Bibliographic Summaries of the Select Literature Volume I", "L'Intelligence Artificielle: promesses et Realities" Humorous Article: A Brief Review of Ignorance Engineering Simon Fraser University AI tech report list ------------------------------ Date: 12 Oct 1985 11:44-CST From: leff%smu.csnet@CSNET-RELAY.ARPA Subject: Canadian Artificial Intelligence 5 Summary of Canadian Artificial Intelligence 5, September 1985 "Canada Prominent in IJCAI Awards" Reports on AIers prominent at IJCAI: Levesque (Computers and Thought), Best Paper to Fagin and Halpern and Professor Randy Goebel who stumped the band in a taping of Mr. Carson's televison show, The Tonight Show. "NSERC Proposes Major Increase in Research Funding" NSERC is the Canadian science funding organization Michel Pilot has started his own AI consulting service Coast Mountain Intelligence specializes in Resource Applications. They completed an expert system on the choice of statistical packages for geophysical data. They are working on expert systems for forest management and interpretation of snow profiles for avalanche predictions Xerox Announces Low Cost Workstations Workshop Report: Theoretical Approaches to Natural Language Understanding Workshop Report: Workshop on the Foundations of Adaptive Information Processing Canada Conquers Los Angeles: mentions Canadians prominent at IJCAI-85 Directory of Candian AI businesses Reviews of: Human Foundations of Advanced Computing Technology: The Guide to the Select Literature from the Report Store Readings in Knowledge Representation Introduction to Artificial Intelligence by Eugene Charniak and Drew McDermott Artificial Intelligence Applications for Business Management Artificial Intelligence Applications for Manufacturing Obituaries for Jeffrey Robert Sampson, Daniel Louis Shalom Berlin Donald Grant Kuehner and David Julian Meredith Davies Tech Report Lists from University of Calgary, University of Montreal, University of Toronto, McGill University, University of Alberta, Simon Fraser University ------------------------------ Date: 14 Oct 1985 12:32-CST From: leff%smu.csnet@CSNET-RELAY.ARPA Subject: AI at conferences International Symposium on Spatial Data Handling at the University of Zurich, August 1984 Order from Symposium Secretariat, Department of Geography, University of Zurich, Winterthurerstrasse 190 CH-8057 Zurich, Switzerland. Price 30 dollars Data Structures for a Knowledge-Based Geographic Information System D. J. Perquet Symbolic Feature Analysis and Expert Systems B. Palmer Autonap -- An Expert System for Automatic Map Name Placement H. Freemand, J. Ahn Knowledge Based Control of Search and Learning in a Large Scale GIS T. R. Smith M. Pazner ____________________________________________________________________________ Graphics Interface 85, 11th Conference of Candian Man-Computer Society Montreal May 27-31 Robotics and CAD/CAM Section Non-rigid Motion A. R. Dill and M. D. Levine McGill University Electronic Assembly by Robots C. Michaud, A. S. Malowany, M. D. Levine McGill University Lo Cost Geometric Modelling System for CAM W. G. Ngai and Y. K . Chan Chinese Univeristy of Hong Kong Panel- Computer Graphics in Environmental Design Artificial Intelligence Generative Design in Architecture Using an Expert System E. Chang University of Victoria Knowledge Engineering Application in Image Processing K. Mikame, N. Sueda, A. Hoshi, S. Ohoniden Toshiba, Japan Visual Perception L. Scholl Laura Scholl and Associates USA ------------------------------ Date: Fri, 11 Oct 85 13:49 EDT From: Kahin@MIT-MULTICS.ARPA Subject: Seminar - Intelligent Electronic Mail (MIT) [Forwarded from the MIT bboard by SASW@MIT-MC.] Massachusetts Institute of Technology Communications Forum Making Electronic Mail More Intelligent October 31, 1985 Thomas Malone, MIT Kenneth Mayers, Digital Equipment Corporation Electronic messaging has become a familiar feature of the office environment and a key element in office automation strategy for many organizations. As these systems spread, many issues must be dealt with, such as accomodating evolving user requirements, responding to rapid expansion, controlling junk mail, and incorporating alternative technologies. One of the central challenges is how to enhance messaging features so that users are not swamped by information overload. This forum will present the experience of Digital Equipment Corporation, one of the pioneering users of electronic mail, and will describe some recent innovative research at MIT which uses artificial intelligence technology to improve the user's ability to sort incoming messages by relevance and urgency and to route outgoing communications to the most appropriate people within the organizations. 4:00 - 6:00 Bartos Theater for the Moving Image The Wiesner Building (Center for Arts and Media Technology) (Building E15 Lower Level) 20 Ames Street Massachusetts Institute of Technology Cambridge, Massachusetts For further information call 617-253-3144. ------------------------------ Date: Mon, 14 Oct 85 11:31:01 EDT From: Bernard Silver Subject: Seminar - Connectionist Learning (GTE) GTE LABORATORIES INC MACHINE LEARNING SEMINAR TITLE: Learning by Statistical Cooperation in Connectionist Networks SPEAKER: Prof. Andrew G. Barto University of Massachusetts at Amherst TIME: 2pm, Wednesday, October 23 PLACE: GTE Laboratories Inc 40 Sylvan Rd Waltham MA 02254 Since the usual approaches to cooperative computation in networks of neuron-like computating elements do not assume that network components have any ``preferences," they do not make substantive contact with game theoretic concepts, despite their use of some of the same terminology. In the approach I describe, however, each network component, or adaptive element, is a self-interested agent that prefers some inputs over others and ``works" toward obtaining the most highly preferred inputs. I describe some of our work with an adaptive element that is robust enough to learn to cooperate with other elements like itself in order to further its self-interests. It is argued that some of the long-standing problems concerning adaptation and learning by networks might be solvable by this form of cooperativity, and computer simulation experiments are described that show how networks of self-interested components that are sufficiently robust can solve rather difficult learning problems. A secondary aim of this talk is to suggest that beyond what I explicitly illustrate, there is a wealth of ideas from game theory and allied disciplines such as mathematical economics that can be of use in thinking about cooperative computation in both nervous systems and man-made systems. For more information contact Bernard Silver (617) 466-2663 ------------------------------ Date: Mon, 14 Oct 85 05:45:36 pdt From: gluck@SU-PSYCH (Mark Gluck) Subject: Seminar - Connectionist Learning (SU) THE COMPUTATION, COGNITION, & NEUROSCIENCE JOURNAL CLUB AND SPORADIC SEMINAR SERIES presents: "From Classical Conditioning to Cognitive Computations" Richard S. Sutton GTE Fundamental Research Laboratory Date: Mon, Oct. 28 Time: 12:00-1:15 Place: Room 100, Jordan Hall One attractive aspect of connectionist models is their ability to make contact with a wide range of fields from neuroscience to cognitive science and AI. In this talk I will review the status of the "connectionist connection" between such fields and present some of my own work as an example of a case in which pursuing it has been fruitful. I will present a sequence of three closely-related connectionist learning models. The first was presented by Sutton and Barto in 1981 as a real-time model of classical conditioning consistent with many behavioral phenomena including blocking, conditioned inhibition, the ISI dependency, higher-order conditioning, and serial-compound effects. The second model is the result of changes made to the first in trying to use it in an AI learning problem. Remarkably, the modified model is not only very effective on the AI problem, but is also a better match to the classical conditioning data than the first model. I am currently working on a third model that is able to reproduce the animal learning phenomena of latent learning and sensory preconditioning. The AI goal for this model is to build a system that can learn about and then reason about its environment. The model shows promise of being simultaneously very successful as all of 1) a model of classical conditioning, 2) an aid to AI machine learning systems, and 3) a model of simple forms of inference and planning. ***************************************************************************** The CCNJCS^3 was formed in response to a growing interest among members of the Psychology, Computer Science, and Neuroscience departments at Stanford in learning about recent advances in the study of computational approaches to modelling the relationship between cognition and neuroscience. In addition to organizing seminars, we also arrange journal club meetings in which graduate students and post-docs meet to read and discuss current research articles dealing with: The neural substrates of learning and memory, computational models of neuronal processes, and the neural bases of cognitive behavior. For more information, contact Mark Gluck (gluck%su-psych@sumex-aim). ------------------------------ Date: Mon 14 Oct 85 07:52:23-PDT From: Ana Haunga Subject: Seminar - Probabilistic Interpretation of Certainty Factors (SU) SIGLUNCH will be held at the Chemistry Gazebo at 12:05-1:00 p.m. Probabilistic Interpretations for MYCIN's Certainty Factors David Heckerman I will show that the original definition of certainty factors (CF's) is inconsistent with the "defining desiderata" of the CF combination functions. I will then show that if this inconsistency is removed by redefining CF's in terms of the desiderata then CF's have probabilistic interpretations. In other words, I will show that certainty factors are nothing more than transformed probabilistic quantities. The construction of the interpretation provides insights into the assumptions made when propagating CF's through an inference net. For example, it can be shown that all evidence which bears directly on a hypothesis must be conditionally independent on the hypothesis and its negation. After presenting the interpretations, I will discuss several ramifications of the correspondence between CF's and probabilities. ------------------------------ End of AIList Digest ******************** From comsat@vpics1 Sat Oct 19 06:18:07 1985 Date: Sat, 19 Oct 85 06:18:03 edt From: comsat@vpics1.VPI To: fox@opus (MILLER,FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: RO Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a002637; 18 Oct 85 18:39 EDT Date: Fri 18 Oct 1985 14:05-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #148 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Sat, 19 Oct 85 06:01 EST AIList Digest Friday, 18 Oct 1985 Volume 3 : Issue 148 Today's Topics: Queries - Go & VMS vs. UNIX for AI & KR Languages for Semantic Nets & Common Lisp Compiler/Interpreter for VAX/750(ULTRIX), Literature - Foreign Language Abstracting, Applications - ALV Demo, AI Tools - YAPS & AI Machines ---------------------------------------------------------------------- Date: 15-Oct-85 22:02-EDT From: David Nicol Subject: the ancient oriental game of Go I am in the process of writing a program to mediate a Go game, and hopefully will be able to write an algorithm or two for playing. If anyone has done any thinking towards algorithms to play Go, or maybe has one already, I would appreciate very much hearing from you. Cscboic@bostonu ------------------------------ Date: Wed, 16 Oct 85 11:50 EST From: "Christopher A. Welty" Subject: VMS vs UNIX for AI development I know this may set people at each others throats, but it is a legitimate concern of mine, so here goes: What experience has there been out there with AI (mainly ES) development on VMS? We use both UNIX and VMS here at RPI, and I have found in my experience that VMS makes it more difficult to do work, and UNIX makes it easier. But there seems to be a number of people (who don't even work for DEC) that swear by VMS. There must be some rational reason for this. I don't really want to see a discussion of the Operating Systems themselves (as that is another Newslist), just what support they offer for development of Expert Systems (mainly LISP, but feel free to add other languages). I know what UNIX offers, let me hear (see) what VMS offers. -Christopher Welty RPI / CIE Systems Mgr. ------------------------------ Date: 16 Sep 85 1615 WEZ From: U02F%CBEBDA3T.BITNET@WISCVM.ARPA (Franklin A. Davis) Subject: Query: Languages for knowledge rep using semantic nets? We are interested in knowledge representation using semantic nets and frames, and we would like to know who has experience with special languages for this purpose. Furthermore, are there distributors for such software packages? Thanks in advance. Regards, Franklin Davis Institut fuer Informatik und angewandte Mathematik Universitaet Bern Laenggassstrasse 51 CH-3012 Bern Switzerland ------------------------------ Date: Thu, 17 Oct 85 9:24:26 EDT From: "Srinivasan Krishnamurthy" <1438@NJIT-EIES.MAILNET> Subject: Common Lisp Compiler/Interpreter for VAX/750(ULTRIX) Dear Readers, Can somebody give me details about a CommonLisp Compiler and interpreter for a VAX/750 running ULTRIX? Heard that KEE and Knowledge Craft(KC) work only on VMS, want to port it to the above enviornment. Any ideas, leads are welcome. Please message me directly at the following net addreses: Mailnet: srini@NJIT-EIES.MAILNET Arpanet: srini%NJIT-EIES.MAILNET@MIT-MULTICS.ARPA USMAIL: S Krishnamurthy COMSAT LABS, NTD. 22300 Comsat Drive. Clarksburg, MD-20871 (301)428-4531 Thanks in advance. Srini. ------------------------------ Date: 17 Oct 1985 0744-PDT (Thursday) From: eugene@AMES-NAS.ARPA (Eugene Miya) Subject: Last call for assistance: helping with foreign language abstracting I would like to thank all of the people who responded for my first call for people to help in the translation/abstraction of foreign language documents. I have been travelling quite a bit during the past five weeks, so next week, I will have a chance to lay the groundwork for determining what journals to monitor and where to post information. For those of you who missed this earlier posting: I am seeking people interested in monitoring foreign language technical documents with an eye to post significant new articles to various bulletin boards. This would be prior to translation, and would hopefully speed translation of potentially significant papers in: AI, graphics, and so forth. Languages which are particularly critical are Eastern Asian: Japanese and Chinese, perhaps French, and other western European languages. We have a few people of each, but it would help to spread the load out. If you are interested, or want to hear more, send me mail to a UUCPnet/ARPAnet gateway listed below. --eugene miya NASA Ames Research Center [Rock of Ages Home for ...] eugene@ames-nas.ARPA UUCP: {ihnp4,hao,hplabs,nsc,cray,research,decwrl}!ames!amelia!eugene ------------------------------ Date: 8 Oct 1985 1205-PDT From: LAWS at SRI-AI.ARPA Subject: Ogling Overseas Technology >From the EE's Tools & Toys column, IEEE Spectrum, Volume 22, No. 10, 10/85, p. 85 The latest research results being published outside the United States may not be as difficult to monitor as one might think. The U.S. Dept. of Commerce publishes a weekly newsletter called Foreign Technology that abstracts new reports and papers from outside the U.S. that are available through the National Technical Information Service. Each abstract lists the report's title, author(s), date, and NTIS publication number, along with a brief synopsis. Topics that the Commerce Department tracks in the newsletter are: biomedical technology; civil, construction, structural, and building engineering; communications; computer technology; electro and optical technology; energy; manufacturing and industrial engineering; materials sciences; physical sciences; transportation technology; and mining and minerals technology. An annual subscription to the weekly newsletter, which is indexed every January, costs $90 in North America. To subscribe or to request subscription prices for other areas, write to U.S. Dept. of Commerce, National Technical Information Service, 5285 Port Royal Rd., Springfield, VA 22161; telephone (703) 487-4630. ------------------------------ Date: 17 Oct 1985 20:51:51 EDT From: Spacely's Space Sprockets Subject: ALV DEMO In response to the recent post regarding the Autonomous Vehicle demo: The Autonomous Land Vehicle project, sponsored by DARPA through the Army ETL, is part of DARPA's Strategic Computing Program, sort of the US's answer to Japan's Fifth Generation effort. Martin Marietta Denver Aerospace ( the Advanced Automation Technology Section ) is the prime contractor of the project -- we are actually developing the system and performing the demos. The project started in late '84 (actually early '85 for most of us), and the first demo was in May '85. We have another demonstration set for next month, and will have about one per year for the next four years, I believe, each demo being more ambitious. The May demo was a preliminary road-following demonstration, with the main point being that we actually got the vehicle going, hardware mounted, some communications figured out, and made it follow a road autonomously. It traversed a 1 km track of road at a speed of about 3 km/h (yep, pretty slow). The vision system (based on a Vicom image processing computer) produced scene models about every 3 seconds for the navigator/pilot to interpret and control the vehicle. The scene model is basically 3D road centerpoints. In November, the vehicle will travel about 5 times as far at speeds up to 10 km/hr and handle things such as shadows on the road, intersecting roads, and sharp curves. We will also be using an ERIM laser range scanner as well as the camera we used in May to provide road images. In later demos we will avoid obstacles, go over cross-country terrain, and other neato tricks. Martin Marietta is officially the integrator of this project, and other universities and companies also have research contracts -- University of Maryland, Carnegie-Mellon University, SRI, AI&DS, Hughes, Honeywell, and maybe some others that I'm not aware of. So far, most of the work contributing to the demos has been done here at Martin Marietta. These other folks will be contributing a lot in the future. A paper describing the project and the May configuration will come out soon in the proceedings of the SPIE Conference on Intelligent Robotics and Computer Vision (Sept. '85) [1]. Matthew Turk MMDA@USC-ISI.arpa [1] Lowrie, Thomas, Gremban, Turk The Autonomous Land Vehicle Preliminary Road-following Demonstration ------------------------------ Date: 18 Oct 1985 09:29-EDT From: Hans.Tallis@ML.RI.CMU.EDU Subject: YAPS Srinivasan, I'm working at Mitre for the summer (tallis@mitre) and we have a version of YAPS which is source-code runnable under Franz, Lambda Zetalisp and Symbolics Zetalisp. Since YAPS is practically public domain, Liz Allen at Maryland probably wouldn't mind our giving you a copy. Send mail if you're intersted. --Hans ------------------------------ Date: 14 Oct 1985 23:48:39-BST From: Aaron Sloman Subject: DO you really need an AI machine? Mon Oct 14 23:47:48 BST 1985 To John and Srinivasan, At Sussex University we have been involved in AI teaching and research for many years. Being British we have a relatively small budget, and for this reason we have resisted going for machines like Symbolics, since a wonderful tool is not much use if you have to spend most of your time queueing up to use it. Instead we have mostly been using VAXen for a range of AI projects. But we did not like the AI software available, so we developed our own - POPLOG.. We've found that with suitable software 10 to 14 AI MSc students can be kept happy most of the time on a 4 Mbyte VAX 750 running Berkeley Unix or VMS. For more advanced researchers the number drops, as it does if you get someone doing image or speech processing. We can support this number because most of the time most people are editing, not running their programs. Of course, we are then stuck with a terrible human-machine interface: a 24 by 80 VDU. So we are now trying to shift as much as possible of our research onto SUN workstations - cheaper than Symbolics. At least SUNs run Unix, unlike most purpose-built AI workstations, and for us that's a big advantage. We use POPLOG, but there's also Quintus Prolog, Common Lisp, and other AI tools available on the SUN. Of course it will be a little while before these machines have the mature interfaces available on Lisp machines. Aaron Sloman, Cognitive Studies Programme, University of Sussex, Brighton, England. ------------------------------ Date: Tue, 15 Oct 85 21:22:55 PDT From: Richard K. Jennings Subject: AI Hype & Big AI Machines. I think Martins (#146) missed the point of Wylands (#145) salient observation that AI researchers focus on *problems* while disciples of other forms of CS focus on *solutions*. For those of you near good college libraries let me suggest that you look up the "Collected Works of John VonNeumann" and read what he has to say about computers. In short, he advocates (see Vol 5) that computers be used to obtain insights into problems, which are then presumably solved in closed form. It makes sense, then, to use AI to develop an understanding of problems which are really too difficult to deal with without AI techniques, and then close in on a CS *solution*, and finally with insights so obtained on a closed form, mathematically verifiable true solution. In my area of interest we have to deal with lots of nasty solutions to differential equations. At last, I have sold my bosses to get Macsyma to help us beat these monster's into a form which we can implement in reasonable time on our mainframes. [Macsyma is a super duper symbolic algebra package which costs 5K for a Symbolics 3670 from Symbolics -- it was developed over the last 20 years at MIT -- and yes, the price just dropped]. With regard to messages from Peck (#144) suggesting small Xerox AI machines and Connolly (#145) and Welty (#144) singing the praises of large AI machines -- there is no doubt that if you have experienced AI investigators, and a network of solid general purpose processing the large AI (and small ones) are worth their cost. The original questions was, however, where does one start? Cugini was correct in pointing out the risks of jumping in to the AI culture too quickly for the reasons he stated and two others: 1) *solutions* often are often dependant upon CS techniques, 2) if you don't know that AI is part of your solution (eg. you are not part of an research AI group) why commit yourself prematurely? Coupled with the availability of good learning systems (and adequate but perhaps not best) development systems on the VAX and PC-AT there is really little need to invest initially in a dedicated AI machine of any size (although Xerox's 6085 sure looks nice). Slow as Golden Hill's Common Lisp is, it is the Lisp system of choice for beginning Lisper's at our organization (we also have a Symbolics 3670 as I implied above). In fact, rolling out of bed in the morning does not qualify one to appreciate the Symbolics development environment. {if it did it probably would not be worth using} [flame off] Richard Jennings Arpa: jennings@aerospace I don't work for them, just use their arpanet port. ------------------------------ Date: Wed, 16 Oct 85 10:21 CDT From: Joseph_Tatem Subject: AI machines I have been reading with interest the discussion about AI (LISP) machines and their usefulness. Since I have been thinking about this myself, I will take this oppurtunity to throw my two cents in. >From what I hear, the average time that it takes to come up to speed on one of these LISP machines is about 3 months. This corresponds roughly to my own experience. Of course, some vendors provide online services that can aid you to various degrees. However, I have found that John Cugini's sentiments are common and not altogether ill-founded. Have you ever done any work with the Window System on one of these beasts?? If you have you know that it is a mess and is not well-documented. You find flavors like STREAM-MIXIN-WITH-HACKS. When you look these up in the manual, you will likely-as-not read something like, "This function does not work reliably, don't use it." On the other hand, once you have learned your way around a little bit, you find that you are using a very powerful machine with a nice development environment. I can get at lots of imformation in the debugger and I can incrementally develop my systems, etc, etc. I find that I have become spoiled. Things that I once did fine without (or with) now seem essential (unnecessary): I don't know how I lived without (with) them before. I believe that the problem is a design issue. Most of these machines are based on software that was developed at (and licensed by) MIT. It seems to me that the system was never really designed (at least the user-interface), but that it was a concatenation of Master's Theses (and other grad. student type work). This is not to say that it is not a good system. There has been a lot of good work put into these machines. It is just that there needs to be some consistency. I see a need to redesign at two levels. First, I would like to see a consistent set of functions (and flavors, etc). Secondly, I would like to see a well-designed user interface. A mouse and a few windows do not make a good interface just by being there. By now, we should know the kinds of things that make computers easy to use. There certainly is no dearth of ideas in the literature. At the very least, I would like for a brand-new user to be able to sit down at one of these beasts and at least be able to figure out which mouse button to click or which function to enter to get himself started. So whaddya think?? Joe Tatem tatem%ti-eg@csnet-relay Note: The opinions expressed herein are strictly my own and in no way reflect those of my employer. ------------------------------ End of AIList Digest ******************** From comsat@vpics1 Sat Oct 19 06:16:29 1985 Date: Sat, 19 Oct 85 06:16:21 edt From: comsat@vpics1.VPI To: fox@opus (MILLER,FRANCE,JOSLIN,ROACH,FOX) Subject: From: AIList Moderator Kenneth Laws Status: RO Received: from sri-ai.arpa by CSNET-RELAY.ARPA id a003564; 18 Oct 85 21:06 EDT Date: Fri 18 Oct 1985 14:16-PDT Reply-to: AIList%sri-ai.arpa@CSNET-RELAY US-Mail: SRI Int., 333 Ravenswood Ave., Menlo Park, CA 94025 Phone: (415) 859-6467 Subject: AIList Digest V3 #149 To: AIList%sri-ai.arpa@CSNET-RELAY Received: from rand-relay by vpi; Sat, 19 Oct 85 06:03 EST AIList Digest Friday, 18 Oct 1985 Volume 3 : Issue 149 Today's Topics: Projects - University of Aberdeen & CSLI, Literature - New Complexity Journal, AI Tools - Lisp vs. Prolog, Opinion - AI Hype & Scaling Up, Cognition & Logic - Modus Ponens, Humor - Dognition ---------------------------------------------------------------------- Date: Thu 17 Oct 85 12:44:41-PDT From: Derek Sleeman Subject: University of Aberdeen Program UNIVERSITY of ABERDEEN Department of Computing Science The University of Aberdeen is now making a sizeable committment to build a research group in Intelligent Systems/Cognitive Science. Following the early work of Ted Elcock and his co-workers, the research work of the Department has been effectively restricted to databases. However, with the recent appointment of Derek Sleeman to the faculty from summer 1986, it is anticipated that a sizeable activity will be (re)established in AI. In particular we are anxious to have a number of visitors at any time - and funds have been set aside for this. So we would be particularly interested to hear from people wishing to spend Sabbaticals, short-term Research fellowships etc. Please contact Derek Sleeman at 415 497 3257 or SLEEMAN@SUMEX for further details. ------------------------------ Date: Wed 16 Oct 85 17:12:46-PDT From: Emma Pease Subject: CSLI Projects [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] CSLI PROJECTS The following is a list of CSLI projects and their coordinators. AFT Lexical Representation Theory. Julius Moravcsik (AFT stands for Aitiuational Frame Theory) Computational Models of Spoken Language. Meg Withgott Discourse, Intention, and Action. Phil Cohen. Embedded Computation Group. Brian Smith (3 sub groups) sub 1: Research on Situated Automata. Stan Rosenschein sub 2: Semantically Rational Computer Languages. Curtis Abbott sub 3: Representation and Reasoning. Brian Smith Finite State Morphology. Lauri Karttunen Foundations of Document Preparation. David Levy. Foundations of Grammar. Lauri Karttunen Grammatical Theory and Discourse Structures. Joan Bresnan Head-driven Phrase Structure Grammar. Ivan Sag and Thomas Wasow Lexical Project. Annie Zaenen Linguistic Approaches to Computer Languages. Hans Uszkoreit Phonology and Phonetics. Paul Kiparsky Rational Agency. Michael Bratman Semantics of Computer Language. Terry Winograd Situation Theory and Situation Semantics (STASS). Jon Barwise Visual Communication. Sandy Pentland In addition, there are some interproject working groups. These include: Situated Engine Company. Jon Barwise and Brian Smith Representation and Modelling. Brian Smith and Terry Winograd ------------------------------ Date: Wed 16 Oct 85 09:56:32-EDT From: Susan A. Maser Subject: NEW JOURNAL JOURNAL OF COMPLEXITY Academic Press Editor: J.F. Traub, Columbia University FOUNDING EDITORIAL BOARD K. Arrow, Stanford University G. Debreu, University of California, Berkeley Z. Galil, Columbia University L. Hurwicz, University of Minnesota J. Kadane, Carnegie-Mellon University R. Karp, University of California, Berkeley S. Kirkpatrick, I.B.M. H.T. Kung, Carnegie-Mellon University M. Rabin, Harvard University and Hebrew University S. Smale, University of California, Berkeley S. Winograd, I.B.M. S. Wolfram, Institute for Advanced Study H. Wozniakowski, Columbia University and University of Warsaw YOU ARE INVITED TO SUBMIT YOUR MAJOR RESEARCH PAPERS TO THE JOURNAL. See below for further information. Publication Information and Rates: Volume 1 (1985), 2 issues, annual institutional subscription rates: In the US and Canada: $60 All other countries: $68 Volume 2 (1986), 4 issues, annual institutional subscription rates: In the US and Canada: $80 All other countries: $93 Send your subscription orders to: Academic Press, Inc. 1250 Sixth Avenue San Diego, CA 92101 (619) 230-1840 Contents of Volume 1, Issue 1: "A 71/60 Theorem for Bin Packing" by Michael R. Garey & David S. Johnson "Monte-Carlo Algorithms for the Planar Multiterminal Network Reliability Problem" by Richard M. Karp & Michael Luby "Memory Requirements for Balanced Computer Architectures" by H.T. Kung "Optimal Algorithms for Image Understanding: Current Status and Future Plans" by D. Lee "Approximation in a Continuous Model of Computing" by K. Mount & S. Reiter "Quasi-GCD Computations" by Arnold Schonhage "Complexity of Approximately Solved Problems" by J.F. Traub "Average Case Optimality" by G.W. Wasilkowski "A Survey of Information-Based Complexity" by H. Wozniakowski SUBMISSION OF PAPERS The JOURNAL OF COMPLEXITY is a multidisciplinary journal which covers complexity as broadly conceived and which publishes research papers containing substantial mathematical results. In the area of computational complexity the focus is on problems which are approximately solved and for which optimal algorithms or lower bound results are available. Papers which provide major new algorithms or make important progress on upper bounds are also welcome. Papers which present average case or probabilistic analyses are especially solicited. Of particular interest are papers involving distributed systems or parallel computers for which only approximate solutions are available. The following is a partial list of topics for which computational complexity results are of interest: applied mathematics, approximate solution of hard problems, approximation theory, control theory, decision theory, design of experiments, distributed computation, image understanding, information theory, mathematical economics, numerical analysis, parallel computation, prediction and estimation, remote sensing, seismology, statistics, stochastic scheduling. In addition to computational complexity the following are among the other complexity topics of interest: physical limits of computation; chaotic behavior and strange attractors; complexity in biological, physical, or artificial systems. Although the emphasis is on research papers, surveys or bibliographies of special merit may also be published. To receive a more complete set of authors' instructions (with format specifications), or to submit a manuscript (four copies please), write to: J.F. Traub, Editor JOURNAL OF COMPLEXITY Department of Computer Science 450 Computer Science Building Columbia University New York, New York 10027 ------------------------------ Date: Tue, 15 Oct 85 22:15 EDT From: Hewitt@MIT-MC.ARPA Subject: Lisp vs. Prolog (reply to Pereira) I would like to reply to Fernando Pereira's message in which he wrote: It is a FACT that no practical Prolog system is written entirely in Lisp: Common, Inter or any other. Fast Prolog systems have been written for Lisp machines (Symbolics, Xerox, LMI) but their performance depends crucially on major microcode support (so much so that the Symbolics implementation, for example, requires additional microstore hardware to run Prolog). The reason for this is simple: No Lisp (nor C, for that matter...) provides the low-level tagged-pointer and stack operations that are critical to Prolog performance. It seems to me that the above argument about Prolog not REALLY being implemented in Lisp is just a quibble. Lisp implementations from the beginning have provided primitive procedures to manipulate the likes of pointers, parts of pointers, invisible pointers, structures, and stack frames. Such primitve procedures are entirely within the spirit and practice of Lisp. Thus it is not surprising to see primitive procedures in the Lisp implementations of interpreters and compilers for Lisp, Micro-Planner, Pascal, Fortran, and Prolog. Before now no one wanted to claim that the interpreters and compilers for these other languages were not written in "Lisp". What changed? On the other hand primitive procedures to manipulate pointers, parts of pointers, invisible pointers, structures, and stack frames are certainly NOT part of Prolog! In FACT no one in the Prolog community even professes to believe that they could EVER construct a commercially viable (i.e. useful for applications) Common Lisp in Prolog. I certainly realize that interesting research has been done using Planner-like and Prolog-like languages. For example Terry Winograd implemented a robot world simulation with limited natural language interaction using Micro-Planner (the implementation by Sussman, Winograd, and Charniak of the design that I published in IJCAI-69). Subsequently Fernando did some interesting natural language research using Prolog. My chief chief concern is that some AILIST readers might be misled by the recent spate of publicity about the "triumph" of Prolog over Lisp. I simply want to point out that the emperor has no clothes. ------------------------------ Date: Thu, 10 Oct 85 11:03:00 GMT From: gcj%qmc-ori.uucp@ucl-cs.arpa Subject: AI hype A comment from Vol 3 # 128:- ``Since AI, by definition, seeks to replicate areas of human cognitive competence...'' This should perhaps be read in the context of the general discussion which has been taking place about `hype'. But it is still slightly off the mark in my opinion. I suppose this all rests on what one means by human cognitive competence. The thought processes which make us human are far removed from the cold logic of algorithms which are the basis for *all* computer software, AI or otherwise. There is an element in all human cognitive processes which derives from the emotional part of our psyche. We reach decisions not only because we `know' that they are right, but also because we `feel' them to be correct. I think really that AI must be seen as an important extension to the thinking process, as a way of augmenting an expert's scope. Gordon Joly (now gcj%qmc-ori@ucl-cs.arpa (formerly gcj%edxa@ucl-cs.arpa ------------------------------ Date: Fri 18 Oct 85 10:13:10-PDT From: WYLAND@SRI-KL.ARPA Subject: Scaling up AI solutions >From: Gary Martins >Subject: Scaling Up >Mr. Wyland seems to think that finding problem solutions which "scale up" >is a matter of manufacturing convenience, or something like that. What >he seems to overlook is that the property of scaling up (to realistic >performance and behavior) is normally OUR ONLY GUARANTEE THAT THE >"SOLUTION" DOES IN FACT EMBODY A CORRECT SET OF PRINCIPLES. [...] The problem of "scaling up" is not that our solutions do not work in the real world, but that we do not have general, universal solutions applicable to all AI problems. This is because we only understand *parts* of the problem at present. We can design solutions for the parts we understand, but cannot design the universal solution until we understand *all* of the problem. Binary vision modules provide sufficient power to be useful in many robot assembly applications, and simple word recognizers provide enough power to be useful in many speech control applications. These are useful, real-world solutions but are not *universal* solutions: they do not "scale up" as universal solutions to all problems of robot assembly or understanding speech, respectively. I agree with you that scientific theories are proven in the lab (or on-the-job) with real world data. The proof of the engineering is in the working. It is just that we have not reached the same level of understanding of intelligence that Newton's Laws provided for mechanics. Dave Wyland ------------------------------ Date: Tue 15 Oct 85 13:48:28-PDT From: Mike Dante Subject: modus ponens Seems to me that McGee is the one guilty of faulty logic. Consider the following example: Suppose a class consists of three people, a 6 ft boy (Tom), a 5 ft girl (Jane), and a 4 ft boy (John). Do you believe the following statements? (1) If the tallest person in the class is a boy, then if the tallest is not Tom, then the tallest will be John. (2) A boy is the tallest person in the class. (3) If the tallest person in the class is not Tom then the tallest person in the class will be John. How many readers believe (1) and (2) imply the truth of (3)? - Mike ------------------------------ Date: Thu, 17 Oct 85 21:22:26 pdt From: cottrell@nprdc.arpa (Gary Cottrell) Subject: Seminar - Parallel Dog Processing SEMINAR Parallel Dog Processing: Explorations in the Nanostructure of Dognition Garrison W. Cottrell Department of Dog Science Condominium Community College of Southern California Recent advances in neural network modelling have led to its application to increasingly more trivial domains. A prominent example of this line of research has been the creation of an entirely new discipline, Dognitive Science[1], bringing together the insights of the previously disparate fields of obedience training, letter carrying, and vivisection on such questions as, "Why are dogs so dense?" or, "How many dogs does it take to change a lightbulb?"[2] This talk will focus on the first question. Early results suggest that the answer lies in the fact that most dog information processing occurs in their brains. Converging data from various fields (see, for example, "A vivisectionist approach to dog sense manipulation", Seligman, 1985) have shown that this "wetware" is composed of a massive number of slow, noisy switching elements, that are too highly connected to form a proper circuit. Further, they appear to be all trying to go off at the same time like popcorn, rather than proceeding in an orderly fashion. Thus it is no surprise to science that they are dumb beasts. Further impedance to intelligent behavior has been discovered by learning researchers. They have found that the connections between the elements have little weights on them, slowing them down even more and interfering with normal processing. Indeed, as the dog grows, so do these weights, until the processing elements are overloaded. Thus it is now clear why you can't teach an old dog new tricks, and also explains why elderly dogs tend to hang their heads. Experience with young dogs appears to bear this out. They seem to have very little weight in their brains, and their behavior is thus much more laissez faire than older dogs. We have applied these constraints to a neural network learning model of the dog brain. To model the noisy signal of the actual dog neurons, the units of the model are restricted to communicate by barking to one another. As these barks are passed from one unit to another, the weights on the units are increased by an amount proportional to the loudness of the bark. Hence we ____________________ [1]A flood of researchers finding Cognitive Science too hard are switching to this exciting new area. It appears that trivial results in this unknown field will beget journal papers and TR's for several years before funding agencies and reviewers catch on. [2]Questions from the Philosophy of dognitive science (dogmat- ics), such as "If a dog barks in the condo complex and I'm not there to hear it, why do the neighbors claim it makes a sound?" are beyond the scope of this talk. term this learning mechanism bark propagation. Since the weights only increase, just as in the normal dog, at asymptote the network has only one stable state, which we term the dead dog state. Our model is validated by the fact that many dogs appear to achieve this state while still breathing. We will demonstrate a live simulation of our model at the talk. ------------------------------ End of AIList Digest ********************