LAWS@SRI-AI.ARPA (04/01/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI> AIList Digest Monday, 1 Apr 1985 Volume 3 : Issue 43 Today's Topics: Expert Systems - EURISKO & DENDRAL/META-DENDRAL & OPS5 for PCs, News - MCC's Bob Inman named to SWB's board of directors, Symbolic Math - Functionals, Meeting - NAIL Journal Club, Seminars - Plans and Situated Actions (UCB) & Models in Syllogistic Reasoning (CSLI) & Programming Descriptive Analogies by Example (MIT) & Functional Role Semantics (CSLI) & NL Understanding and Generation (CMU) ---------------------------------------------------------------------- Date: Sat 30 Mar 85 18:18:45-PST From: Lee Altenberg <ALTENBERG@SUMEX-AIM.ARPA> Subject: EURISKO Does anyone know who in the Stanford area is actively working with EURISKO, if anyone, now that Doug Lenat is in Texas? ------------------------------ Date: Mon, 1 Apr 85 09:02:48 mst From: cib@LANL.ARPA (C.I. Browne) Subject: DENDRAL/META-DENDRAL Can anyone tell me where either source or object code for DENDRAL and META-DENDRAL may be obtained? SUMEX advises that they no longer either support or distribute the program, and refer me to Molecular Design Limited. While that company markets several interesting programs, neither DENDRAL nor META-DENDRAL is among them. It would be a pity if such programs have disappeared from the scene and become unavailable to sites wishing to include them in an AI library. Thank you. cib ------------------------------ Date: 30 March 1985 1049-EST From: Peter Pirolli@CMU-CS-A Subject: OPS5 for PCs I just received a flyer for a language called TOPSI which is supposed to be an OPS5 clone developed for CP/M and MS-DOS machines. Here are some quotes from the flyer: "TOPSI has the full power of the original language PLUS extensions to improve its computational and list management capabilities. Usage of memory and computer time is optimized for use on home computers." "TOPSI rule are written in a simple, legible form and compiled into a memory-efficient data structure enabling fast execution." "TOPSI's rule base and data base can be saved separately allowing a system to be exercised easilty with different data sets." "TOPSI is available on 5 1/4 in diskettes for 65k CP/M systems or MS-DOS with at least 128k of memory. It comes complete with a users manual, example programs, a tutorial section on writing your own production systems, and a 30 day warranty." The company is: Dynamic Master Systems Inc. P.O. Box 566456 Atlanta, GA 30356 [404] 565-0771 ------------------------------ Date: Sat 30 Mar 85 15:21:43-CST From: Werner Uhrig <CMP.WERNER@UTEXAS-20.ARPA> Subject: MCC's Bob Inman named to SWB's board of directors [ from the Austin American Statesman - March 30, 1985 ] Bob Inman, chairman and chief executive officer of MCC has been elected to the board of directors of Southwestern Bell Corp. ------------------------------ Date: Wed, 20 Mar 85 09:48:24 PST From: "David G. Cantor" <DGC@UCLA-LOCUS.ARPA> Subject: If f(f(x)) = x^2 - 2, what is f(x)? [Forwarded from the Prolog Digest by Laws@SRI-AI.] Q: "Can a computer solve the query: "If f(f(x)) = x^2 - 2, what is f(x)? If so, how?" [This question was forwarded to the Prolog Digest a few weeks ago by Nils Nilsson. -- KIL] The solution is essentially contained in the article by Michael Restivo in the March 20 issue of Prolog Digest. The nth Tchebycheff Polynomial may be defined as n n T (x) = u + v , n where u = (x + d)/2, v = (x - d)/2, with d = sqrt(x * x - 4). It is easy to check that, when n is an integer, the powers of d cancel and hence that the above functions are really are polynomials. These polynomials satisify numerous identities. The pertinent one here is that T (T (x)) = T (x) . m n m * n This can be verified by elementary algebra (note that u * v = 1). It holds certainly for all complex numbers m and n, subject to choosing appropriate branches of the mth and nth power as well as the square root, in the complex plane. The function f(x) = T (x) sqrt(2) then satisfies f(f(x)) = T (T (x)) sqrt(2) sqrt(2) = T (x) sqrt(2) * sqrt(2) = T (x) 2 = x * x - 2, and hence solves the original problem. As to how a computer could solve this: It need only search the mathematical literature to find a paper by Michael Fried giving all solutions to the functional equation (due to Issai Schur): F (F (x)) = F (x) . m n m * n Fried shows that, under very general conditions, the solutions are either n F (x) = x or F (x) = T (x) , n n n as given above. The computer then need only recognize that the given function f(x) = T (x) . 2 Alternatively it could recognize the latter first, and be led to study identities of the Tchebycheff polynomials. -- David G. Cantor ------------------------------ Date: Fri, 29 Mar 85 12:13:39 pst From: Jeff Ullman <ullman@diablo> Subject: NAIL! (Not another implementation of Logic!) Journal Club [Forwarded from the Stanford BBoard by Laws@SRI-AI.] There is a meeting of people interested in implementation of database systems with a "knowledge" component, i.e., a logical language providing access to a database. We primarily read and present papers, and present our own ideas on the subject. The first Spring meeting is 1PM Weds. 4/3, in 252MJH, and subsequent meetings are Wednesdays, 11AM in 301MJH. You can get on the nail list by mailing to mailer@diablo a message with *subject heading* add <yourname> to nail ------------------------------ Date: Thu, 28 Mar 85 14:45:45 pst From: chertok%ucbcogsci@Berkeley (Paula Chertok) Subject: Seminar - Plans and Situated Actions (UCB) BERKELEY COGNITIVE SCIENCE PROGRAM Cognitive Science Seminar -- IDS 237B TIME: Tuesday, April 2, 11 - 12:30 PLACE: 240 Bechtel Engineering Center (followed by) DISCUSSION: 12:30 - 1:30 in 200 Building T-4 SPEAKER: Lucy Suchman, Intelligent Systems Laboratory, Xerox PARC TITLE: ``Plans and Situated Actions: the problem of human-machine communication'' Researchers in Cognitive Science view the organization and significance of action as derived from plans, which are prere- quisite to and prescribe action at whatever level of detail one might imagine. Mutual intelligibility on this view is a matter of the recognizability of plans, due to common conventions for the expression of intent, and common knowledge about typical situations and appropriate actions. An alternative view, drawn from recent work in social science, treats plans as derived from situated actions. Situated actions as such comprise necessarily ad hoc responses to the actions of others and to the contingencies of particular situations. Rather than depend upon the reliable recognition of intent, successful interaction consists in the collaborative production of intelligibility through mutual access to situation resources, and through the detection, repair or exploitation of differences in understand- ing. As common sense formulations designed to accomodate the unforseeable contingencies of situated action, plans are inherently vague. Researchers interested in machine intelli- gence attempt to remedy the vagueness of plans, to make them the basis for artifacts intended to embody intelligent behavior, including the ability to interact with their human users. I examine the problem of human-machine interaction through a case study of people using a machine designed on the planning model, and intended to be intelligent and interactive. A conversation analysis of "interactions" between users and the machine reveals that the machine's insensitivity to particular circumstances is both a central design resource, and a funda- mental limitation. I conclude that problems in Cognitive Science's theorizing about purposeful action as a basis for machine intelligence are due to the project of substituting plans for actions, and representations of the situation of action for action's actual circumstances. ------------------------------ Date: Wed 27 Mar 85 17:18:44-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Seminar - Models in Syllogistic Reasoning (CSLI) [Excerpted from the CSLI newsletter by Laws@SRI-AI.] CSLI ACTIVITIES FOR *NEXT* THURSDAY, April 4, 1985 2:15 p.m. CSLI Seminar Redwood Hall ``Manipulating Models in Syllogistic Reasoning'' Room G-19 Marilyn Ford, CSLI Discussion leader to be announced Johnson-Laird has argued that reasoners do not use formed rules of inference in solving problems involving syllogistic reasoning, but rather that they come to a solution by manipulating mental models. I will show that while this certainly appears to be true, a number of details of Johnson-Laird's theory appear to be incorrect. An alternative theory will be presented. --Marilyn Ford ------------------------------ Date: 28 Mar 1985 12:06 EST (Thu) From: "Daniel S. Weld" <WELD%MIT-OZ@MIT-MC.ARPA> Subject: Seminar - Programming Descriptive Analogies by Example (MIT) [Forwarded from the MIT bboard by SASW@MIT-MC.] AI REVOLVING SEMINAR Programming Descriptive Analogies By Example Henry Lieberman Before making programs that can perform analogies by themselves, we can attack the more modest goal of being able to communicate to the computer an analogy which is already understood by a person. I will describe a system for "programming by analogy", called Likewise. This new approach to interactive knowledge acquisition works by presenting specific examples and pointing out what aspects of the examples illustrate the more general case. The system constructs a general rule which abstracts out the important aspects so the rule can be applied to "analogous" examples. Given a new example, the system can then construct an analogy with the old example by trying to instantiate analogous descriptions which correspond to the descriptions constructed for the first example. If a new example doesn't fit an old concept exactly, a concept can be generalized or specialized incrementally to make the analogy go through. The operation of the analogy system on a typical concept learning task is presented in detail. Tuesday April 2, 1985 4:00pm 8th floor playroom ------------------------------ Date: Wed 27 Mar 85 17:18:44-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Seminar - Functional Role Semantics (CSLI) [Excerpted from the CSLI newsletter by Laws@SRI-AI.] CSLI ACTIVITIES FOR *NEXT* THURSDAY, April 4, 1985 4:15 p.m. CSLI Colloquium Redwood Hall ``Two Cheers for Functional Role Semantics'' Room G-19 Ned Block, Massachusetts Institute of Technology There are two quite different frameworks for semantics: REDUCTIONIST approaches attempt to characterize the semantic in non-semantic terms. NON-REDUCTIONIST approaches are more concerned with relations among meaningS than with the nature of meaning itself. The non-reductionist approaches are the more familiar ones (eg., Montague, the model-theoretic aspect of situation semantics, Davidson, Katz). The reductionist approaches come in 4 major categories: 1. Theories that reduce meaning to the mental. (This is what is common to Grice and Searle.) 2. Causal semantics--theories that see semantic values as derived from causal chains leading from the world to our words. 3. Indicator semantics--theories that see natural and non-natural meaning as importantly similar. The paradigm of meaning is the way the rings on the tree stump represent the age of the tree when cut down. (Dretske/Stampe, and, in my view, though not in Barwise and Perry's, Situations and Attitudes) 4. Functional role semantics--theories that see meaning in terms of the functional role of linguistic expressions in thought, reasoning, and planning, and in general in the way they mediate between sensory inputs and behavioral outputs. After sketching the difference between the reductionist and non- reductionist approaches, I will focus on functional role semantics, a view that has independently arisen in philosophy (where its sources are Wittgenstein's idea of meaning as use, and pragmatism) and cognitive science (where it is known as procedural semantics). I will concentrate on what theories in this framework can DO, e.g., illuminate acquisition of and knowledge of meaning, principles of charity, how meaning is relevant to explanation of behavior, the intrinsic/observer-relative distinction, the relation between meaning and the brain, and the relativity of meaning to representational system. The point is to give a sense of the fertility and power of the view, and so to provide a rationale for working on solutions to its problems. Finally, I will sketch some reasons to prefer functional role semantics to the other reductionist theories. A copy of a paper which the talk draws on will be in the Ventura reading room. --Ned Block ------------------------------ Date: 29 March 1985 1352-EST From: Theona Stefanis@CMU-CS-A Subject: Seminar - NL Understanding and Generation (CMU) [Forwarded from the CMU bboard by Laws@SRI-AI.] Name: NL Seminar Date: 3 April Time: 11:00-12:00 Place: WeH 7220 The Janus System: Coordinating Understanding and Generation Norman K. Sondheimer USC/Information Science Institute Technology for natural language understanding and generation differs significantly. In cases where they have both been employed in the same system, the results have been an impression of a system that could not understand what it could say. As part of ARPA's Strategic Computing Initiative, researchers at USC/Information Sciences Institute and Bolt Beranek and Newman have begun the design of a system that may be able to avoid these problem. The system employs the ATN parsing, KL-ONE based semantic interpretation, and the NIGEL systemic grammar generator. Much of the integration of understanding and generation will come from a large domain knowledge base developed in the NIKL (New Implementation of KL-One) knowledge representation language. This talk will be a short, informal look at the goals of the effort and the system's initial design. ------------------------------ End of AIList Digest ********************