LAWS@SRI-AI.ARPA (03/04/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI.ARPA> AIList Digest Monday, 4 Mar 1985 Volume 3 : Issue 28 Today's Topics: Knowledge Representation - Attribution of Characteristics & RETE Algorithm & Commonsense and Qualitative Reasoning, AI Tools - KEE Unification & XLISP, Seminar Summary - Representational Cognitive Modeling (CSLI), Seminars - Varieties of Phenomenology (UCB) & Prolog, Databases, and Natural Language Access (SU) & Connectionist Inference Architecture (CMU) & Animating Programs Using Smalltalk (GE) ---------------------------------------------------------------------- Date: 1 Mar 85 07:50:00 EST From: bogner@ari-hq1 Reply-to: <bogner@ari-hq1> Subject: ATTRIBUTION OF CHARACTERISTICS I AM COLLECTING DATA FOR A KNOWLEDGE REPRESENTATION MODEL, ON THE ATTRIBUTION OF CHARACTERISTICS TO INDIVIDUALS USING A MODIFICATION OF GEORGE KELLY'S REP GRID. I AM DEVELOPING AN ARGUMENT AGAINST ATTRIBUTION ALONG BIPOLAR DIMENSIONS (BOUNDED BY ANTONYMS). IS ANYONE ELSE ADDRESSING THAT OR A SIMILAR QUESTION??? I WOULD GREATLY APPRECIATE LISTINGS OF ANY RELATED CONFERENCES, MEETINGS, TUTORIALS, . . . SUE ------------------------------ Date: 1 Mar 1985 9:08:28 EST (Friday) From: Karl Schwamb <m13820@mitre> Subject: RETE Algorithm This is in replay to Don Rose's message to the AIList (V 3, N 27). The two references below are an excellent place to start learning about the RETE algorithm: C. L. Forgy, "Rete: A Fast Algorithm for the Many Pattern/Many Object Pattern Match Problem," in ARTIFICIAL INTELLIGENCE (V 19, N 1). C. L. Forgy, On the efficient implementation of production systems, Ph.D. Thesis, Carnegie-Mellon Univ., 1979. Hope this helps! ...Karl ------------------------------ Date: 1 Mar 85 08:20 PST From: Bobrow.pa@XEROX.ARPA Subject: Commonsense Reasoning From: arora@buffalo (Kulbir S. Arora) Subject: Request for bibliography Is there a bibliography available on Common-sense reasoning systems (qualitative reasoning, mental models) ? Kulbir Arora Volume 24 of the AI Journal was devoted to Qualitative Reasoning about Physical Systems, and is now avalailable as a book from MIT Press. danny bobrow ------------------------------ Date: Fri, 1 Mar 85 17:12 EST From: Paul Fishwick <Fishwick%upenn.csnet@csnet-relay.arpa> Subject: Qualitative Reasoning As to the request for a bibiliography for qualitative reasoning, I suggest reviewing the following 2 references: 1) "Mental Models", edited by Gentner, Dedre and Stevens, Albert, Lawrence Erlbaum Associates, Hillsdale, New Jersey 1983. 2) Special Issue of the Artificial Intelligence Journal: Volume 24, numbers 1-3, December 1984. They contain a number of very good papers in addition to further references for more specific topics within QR. -paul ------------------------------ Date: Sun 3 Mar 85 08:49:32-PST From: Richard Fikes <FIKES@USC-ECL.ARPA> Subject: Rule Systems Using Unification Regarding your query about production rule systems and unification -- The rule system in IntelliCorp's KEE system (Release 2.0) uses full unification to match rules to goals, subgoals, and items retrieved from the knowledge base. It has both a forward chainer and a backward chainer. richard fikes (FIKES@USC-ECL) ------------------------------ Date: Friday, 1 Mar 1985 06:37:27-PST From: minow%rex.DEC@decwrl.ARPA Subject: XLISP Re: XLISP question in AIList V3.27 There is an article on XLISP in the March 85 Byte magazine. Also, the recent USENET distribution of XLISP mentioned that the author, Dave Betz, can be reached as "harvard!betz". I don't know the USENET path to harvard, nor do I know if betz@harvard.arpa would work. Dave's a good guy, please don't bother him with questions if you haven't read the article (and the code/documentation/examples). Martin Minow minow%rex.dec@decwrl.arpa ------------------------------ Date: 1 Mar 1985 1656 PST From: Larry Carroll <LARRY@JPL-VLSI.ARPA> Reply-to: LARRY@JPL-VLSI.ARPA Subject: XLISP The latest Byte (March '85) has an article by the author of XLISP, David Betz. It includes a fairly clear explanation of the concepts you're having trouble with and some examples. Larry @ jpl-vlsi ------------------------------ Date: Fri, 1 Mar 85 18:28:31 PST From: Richard K. Jennings <jennings@AEROSPACE.ARPA> Subject: XLISP 1.4 The current issue of Byte (March 85) has an excellent article on XLISP by its creator David Betz. In it he mentions version 1.4 which is supposed to be similar to Common Lisp. Anybody know where a copy can be obtained? Rich. ------------------------------ Date: Sun, 3 Mar 85 16:00:23 PST From: Richard K. Jennings <jennings@AEROSPACE.ARPA> Subject: XLISP I am in the process of getting the newest version of XLISP to run on an IBM-PC-AT/DOS 3.0/Computer Innovations ver 2.1/ big-dos2-soft model. The source is available at HARVARD, which accepts an anonymous login. In order to get the thing to recompile (almost 1 hr), some changes had to be made, described below. Although I have just started to test it out, any obvious mistakes I may have made I would appreciate hearing about from an observant reader. [XLISP ver 1.4 is a public domain version of a COMMON LISP, new with version 1.4, which was authored and maintained by David Betz] [The source is 170K, .EXE is 70K, and the manual is 30p] [Also see March 85 BYTE for an article by Betz on XLISP] Changes: 1) NIL => 0 from (NODE *) 0 2) deleted references to xlintern() 3) used standard unix "setjmp.h" (bsd 4.1) 4) used a dummy "ctype.h" The file available on HARVARD is a shell file. Since I am not running under unix, I had to kludge my own decommutator which I can send upon request. Richard K Jennings AF Sat Cntl Fac 408 744-6427 av: 799-6427 arpa: jennings@aerospace afscf.xrp@hq-afsc ------------------------------ Date: Thu 21 Feb 85 14:50:16-PST From: Emma Pease <Emma@SU-CSLI.ARPA> Subject: Seminar Summary - Representational Cognitive Modeling (CSLI) [Excerpted from the CSLI Newsletter by Laws@SRI-AI.] SUMMARY OF F4 MEETING At the meeting of project F4 on February 11, Bob Moore presented arguments for the representational approach to designing AI systems and modelling mental activities in humans. Moore first noted the relative ease with which a human can acquire individual beliefs without disturbing very much of the rest of his mental state. This supports the idea that distinct beliefs ought to be embodied more-or-less individually, since acquiring a new belief does not seem to require wholesale reorganization of one's mental state. Moore went on to argue that the combinatorial structure of what can be believed suggests a similar combinatorial structure to how it is believed. The idea is that the combinatorial structure of the sentences used to characterize belief states does not serve merely to distinguish one belief state from another; there are regularities in behavior that depend on that structure. For instance, having a belief of the form ``if not P, then Q'' is associated with behavior appropriate to Q's being true when evidence of P's being false is presented, but not necessarily with behavior appropriate to P's being true when evidence of Q's being false is presented, even though ``if not P, then Q'' and ``if not Q, then P'' are equivalent under most interpretations of the conditional. The fact that this and many other structural distinctions in sentences used to classify belief states correspond to systematic distinctions in behavior presents a prima facie case that the belief states themselves are similarly structured. But, Moore argued, under a conception of representation sufficiently abstract to cover the kinds of ``representation'' actually used in computational models of mental states, the claim that mental states involve ``syntactic'' representations--a language of thought--probably comes to no more than this. Moore concluded by noting that none of these arguments bear on the question of whether the language of thought is distinct from natural language, but that empirical considerations, such as the indexicality of natural language and the difficulty of stating principles of reasoning that apply directly to natural language, suggest that the two are distinct. ------------------------------ Date: Wed, 27 Feb 85 16:15:29 pst From: chertok%ucbkim@Berkeley (Paula Chertok) Subject: Seminar - Varieties of Phenomenology (UCB) BERKELEY COGNITIVE SCIENCE PROGRAM Spring 1985 Cognitive Science Seminar -- IDS 237B TIME: Tuesday, March 5, 11 - 12:30 PLACE: 240 Bechtel Engineering Center (followed by) DISCUSSION: 12:30 - 2 in 200 Building T-4 SPEAKER: Hubert Dreyfus, Department of Philosophy, UC Berkeley TITLE: ``Varieties of Phenomenology: Husserl, Heidegger and Merleau-Ponty'' A tutorial review of the three most important accounts of intentionality in recent continental philosophy, with emphasis on their relevance to current theories of mental representation. Edmund Husserl begins the phenomenological concern with inten- tionality. In his earlier work, The LOGICAL INVESTIGATIONS, he holds a view similar to Searle's that intentional content type individuates mental acts. Later, in IDEAS, he changes to a posi- tion, which he calls ``cognitive science,'' in which mental representations are held to be hierarchies of strict rules, involved in all intelligent activity. I take this to be an early version of the computational view of the mind. Husserl's account leads to two important counter-views. Martin Heidegger in BEING AND TIME argues that intentional states do not play the central role in intelligent behavior Husserl sup- posed, and that even in those cases where intentional states are involved their intentional content can not be treated as abstract structures. Maurice Merleau-Ponty, like Heidegger, argues for a primitive form of intentionality which does not involve mental representation, but whereas Heidegger is primarily interested in an account of action and its social setting, Merleau-Ponty bases his critique on a phenomenology of perception and bodily skills. Together, Heidegger's and Merleau-Ponty's work constitutes the most powerful critique of cognitivism so far offered. ------------------------------ Date: Thu 28 Feb 85 09:51:18-PST From: Gio Wiederhold <WIEDERHOLD@SU-SCORE.ARPA> Subject: Seminar - PROLOG, DATABASES, AND NATURAL LANGUAGE ACCESS (SU) [Forwarded from the Stanford bboard by Laws@SRI-AI.] CS 300 -- Computer Science Department Colloquium -- Winter 1984-1985. Tuesday, March 5, 1985 at 4:15 in Terman Auditorium PROLOG, DATABASES, AND NATURAL LANGUAGE ACCESS David H.D. WARREN Quintus Computer Systems, Inc. PROLOG is a general purpose programming language based on logic. It can be viewed either as an extension of pure LISP, or as an extension of a relational database query language. It was first conceived in 1972, by Alain Colmerauer at the University of Marseille. Since then, it has been used in a wide variety of applications, including natural language processing, algebraic symbol manipulation, compiler writing, architectural design, VLSI circuit design, and expert systems. PROLOG was chosen as the initial kernel language for Japan's Fifth Generation Computer Systems project, and the project's prototype Prolog machine, PSI, has recently been unveiled in Tokyo. In this talk, I will give an overview of the language, and then focus on one particular application, a domain-independent system for natural language question answering, called "CHAT". I will compare the way Chat plans and executes a query with the query optimization strategies of relational database systems such as SYSTEM-R. Finally I will discuss the future prospects for PROLOG in the light of Japan's Fifth Generation project, and describe the high-performance PROLOG systems for the SUN and VAX available from Quintus. ------------------------------ Date: Friday, 1 March 1985 15:56:29 EST From: Steven.Shafer@cmu-cs-ius.arpa Subject: Seminar - Connectionist Inference Architecture (CMU) AI SEMINAR Symbols Among the Neurons: Details of a Connectionist Inference Architecture Dave Touretzky, CMU Tuesday, March 5, 3:00 pm in WeH 5409 Pattern matching and variable binding are easily implemented in conventional computer architectures, but not necessarily in all architectures. In a distributed neural network architecture each symbol is represented by activity in many units and each unit contributes to the representation of many symbols. Manipulating symbols using this type of representation is not as easy as with a local representation where each unit denotes one symbol, but there is evidence that the distributed approach is the one chosen by nature. In this talk I will describe work I am doing with Geoff Hinton on production system interpreters implemented in neural networks using distributed representations for both symbols and rules. The research provides an account of two important symbolic reasoning operations, pattern matching and variable binding, as emergent properties of collections of neuron-like elements. The success of our production system implementations goes some way towards answering a common criticism of connectionist theories: that they aren't powerful enough to do symbolic reasoning. ------------------------------ Date: Tue, 26 Feb 85 15:13:02 EST From: coopercc@GE-CRD Subject: Seminar - Animating Programs Using Smalltalk (GE) Computer Science Seminar General Electric R & D Center Schenectady, N.Y. Animating Programs Using Smalltalk Ralph L. London Tektronix, Inc. Friday, March 15 1:30 PM, Bldg. K1, Conf. Rm. 2 (Refreshments at 1:15) ABSTRACT: We discuss our work in program animation using the Smalltalk programming environment. We strive to isolate the graphical viewing structure from the code of the algorithm being animated. In addition to "procedure calls" this is achieved through a refinement of the Smalltalk Model-View-Controller construct and view dependency mechanism, which allows the algorithm code to broadcast interesting events and supports the insertion of probes into active values. Multiple, dif- ferent views of a single object are easily achieved. There are connections between interesting events and invariant assertions. Further efforts were made to ensure smoothness of motion and transitions between states. A number of particular animations, the evolu- tion of an animation, and directions for further research are included. We plan to show a videotape. Notice to Non-GE attendees: It is necessary that we ask you to notify Marion White (518-385-8370 or WHITEMM@GE-CRD) at least two days in advance of the seminar. ------------------------------ End of AIList Digest ********************