LAWS@SRI-AI.ARPA (04/25/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI> AIList Digest Thursday, 25 Apr 1985 Volume 3 : Issue 53 Today's Topics: Request - MRS Information, Applications - Architecture & Agricultural AI, Psychology - Emotional Attachment & Semantics of Humor, Philosophy - Knowledge, Information, and Belief & Knowledge as an Obstacle to Learning, Application & Humor - BBoard Contest, Seminars - Bertrand Constraint Language (SRI) & Assumption-Based Truth Maintenance (SU) ---------------------------------------------------------------------- Date: 23 Apr 85 15:52:40 PST (Tue) From: whiting@sri-spam Subject: Request for additional information on MRS I am looking at making some extensions/improvements to MRS. The following are being investigated: - Adding a better user-interface - Improving efficiency - Adding debugging aids - Extending the system to deal with uncertainty elegantly I have all the HPP reports on MRS, I am looking for additional information. If you have worked with the system or know of articles which might be useful to me, I would very much appreciate hearing from you. Thanx in Advance, Kevin Whiting (415) 859-4099 ------------------------------ Date: 24 Apr 1985 1119-EST From: Benoit Flamant <FLAMANT@CMU-CS-PS2.ARPA> Reply-to: FLAMANT@CMU-CS-PS2.ARPA Subject: Expert Systems in Architecture I am looking for information about experts systems in the domain of architecture. Does anyone know of good articles or books published recently, understandable by people without deep knowledge in AI ? Please reply to flamant@cmu-cs-ps2. Thanks. ------------------------------ Date: Sat 20 Apr 85 17:42:22-PST From: LOUROBINSON@SRI-AI.ARPA Subject: Agricultural AI In response to Peter Friedland's inquiry regarding agricultural AI systems: The Imperial Chemical Industries of Great Britain recently entered into an agreement with a British expert systems company, ISIS Systems, to develop tools that will assist farmers. One of these tools is a system called Counselor intended to help farmers analyze crop diseases. The advisory service calculates the probable incidence of disease based on data provided by the user. "Wheat Counselor" can be used to determine probable wheat infestations and appropriate (chemical) treatment. Lou Robinson The AI Report ------------------------------ Date: 20 Apr 85 17:16 EST (Sat) From: _Bob <Carter@RUTGERS.ARPA> Subject: Midnight Theorizer From: BATALI%MIT-OZ at MIT-MC.ARPA From: MINSKY Perhaps this, too, explains the prolonged, mourning-like depression that follows sexual or other forms of personal assault. No matter that the unwelcome intimacy of violence may be brief; it nonetheless affects one's attachment machinery, however much against one's wish. So the suggestion is that the rape-victim feels bad because she has formed an attachment-bond to her attacker? The same "mechanism" is involved as in the formation of her attachment-bonds to other people? That is precisely the suggestion. So she feels bad not because she has been raped, but because her rapist has then left her? Is there a shred of evidence that any rape victim has ever felt this way? No, perhaps the victim feels bad because of a. The terrible invasion of the sexual assault itself, b. The cognitive dissonance when the attachment machinery begins to operate in these terribly inappropriate circumstances. I suspect that many people would agree there is more than a shred of evidence for similar emotive patterns. The syndrome in which hostages identify with their captors, despite violence (and in the case of Patty Hearst, rape) comes to mind. Is this theory somehow suggesting that there really isn't much of a difference between rape and seduction and falling in love? Perhaps not, in some ways. Is it being assumed that fear, pain and loss of self-esteem is not enough to "explain the prolonged depression" that follows sexual assault? No, rather it seems to be examining the internal dynamics of "loss of self-esteem." I have some doubts about that part of AI which asserts the validity of scientific inquiry that gathers "data" by introspection. But I have even greater doubts about moral indignation as a criterion for rejection of hypotheses (or of humor, for that matter). _B ------------------------------ Date: Sun, 21 Apr 85 11:51 EST From: Brant Cheikes <Brant%upenn.csnet@csnet-relay.arpa> Subject: Book on semantics of humor Since there was some recent mention on studies of humor, I thought I'd mention a book I just started reading on that very topic: "Semantic Mechanisms of Humor" Raskin, Victor D. Reidel Publishing Company Dordrecht, Holland, 1985. Raskin attempts to develop a semantic theory that captures the necessary and sufficient conditions for a text to be considered funny by the "native speaker." The book is full of examples, however, in the preface, Raskin insists that the book is not a joke book, rather, that all joke examples were chosen purely for their illustrative value. So if you want to know what's funny and what's not and why, then read this book. Brant ------------------------------ Date: Mon, 22 Apr 85 11:04:49 EST From: Morton A Hirschberg <mort@BRL-BMD.ARPA> Subject: Re: Jeff Peck Jeff's got a good start on the data, information, knowledge hierarchy (you see it here as you read from left to right, think from bottom to top). Data is all around us and can exist without context. Information is derived from data in a particular context with or without the aid of data, information, and knowledge from other contexts. From information we can derive secondary data. Think of the two triplets of numbers representing the coordinates of two points in three space. Each by itself is meaningless however, assuming that the triplets are not identical, one can derive the equation of a line in space. From that equation, one can then derive as many new triplets as one desires (some may actually be useful). Knowledge can be derived from information, that is it is the synthesis of information (usually thought of in particular contexts). Thinking about the hierarchy in this way has been quite useful in dealing with and teaching data basing. We also see that data which can not be manipulated to form information remains data. The listing of names, street numbers and telephone numbers in a telephone book is really data and those books should be called data books and not information books. There are ways of manipulating those data (I don't work for Ma Bell) to derive information (Ma Bell sells those books to businesses) but that is another story. Mort ------------------------------ Date: Mon, 22 Apr 85 10:36:39 EST From: cugini@NBS-VMS Subject: definition of knowledge/information/data Alright, my philosophical feathers have been ruffled... "True justified belief" is *not* a special case of "things that are useful in making good decisions". 1. While it probably is true as a matter of empirical fact that true beliefs are likely to be more useful for making decisions than are false ones (in most cases), this is certainly not true by definition. 2. One quick counter-example: Suppose you were in a prisoner-of-war camp. You might have a false, unjustified belief that you would soon be rescued. This irrational hope might in fact have greater survival value (= lead to better decisions on how to act) than a more realistic outlook - but the utility of the belief hardly makes it true, in the any normal English sense of the word. 3. Anyway, the emphasis of the "true justified belief" concept is on "justified" - ie defining knowledge as a true belief rationally arrived at, rather than as a lucky guess. There are problems with this defintion, to be sure, in particular, distinguishing between subjective justification (I believe X, based on the evidence given to me so far), vs objective (but in fact part of that evidence consists of lies, undetected by you). A typical case discussed is: suppose Mr. X wrongly believes that P, and, wanting to deceive you, tells you that not-P is the case (which in fact it is). You then, believing him, have a true belief that not-P is so. But is it justified? Subjectively, yes, because you are acting rationally - but in fact you're relying on a method (Mr. X) which is unreliable - if P really were the case, this method would not lead you to believe P - so your source of belief is not related to its object in such a way that it would reliably track that object in other "nearby" possible worlds. Robert Nozick in "Philosophical Questions" has a very well-written insightful discussion on all this. John Cugini <Cugini@NBS-VMS> National Bureau of Standards Bldg 225 Room A-265 Gaithersburg, MD 20899 phone: (301) 921-2431 ------------------------------ Date: Wed, 17 Apr 85 11:24:58 pst From: Vaughan Pratt <pratt@Navajo> Subject: Knowledge as an obstacle [Forwarded from the Stanford bboard by Laws@SRI-AI.] This morning I overhead "I just don't understand Karl Marx" as I biked past two students. My first reaction was "Gee, if they think Marx is tough they must go bananas with Kant." Then on further reflection I thought "Now if Marx is really that inaccessible, where did he get such a large following?" Which train of thought led me to: There are two very different reasons for being unable to grasp an idea. The "obvious" one is that it takes time to assimilate new ideas. The less obvious one is that you may already know too much! If the new idea is inconsistent with what you know, something has to give. Now consider the following scenario. A teacher of Marx spends say forty years expounding Marx's ideas. In his youth he has no trouble getting the message across. Later he finds it much harder. He agonizes: "Are students really getting dumber, or am I just losing my touch?" The truth of the matter may be neither. He and his students may simply have diverging theories of the world. His stays put while theirs (collectively!) continue to accumulate the latest ideas. The more divergent those theories become, the harder he finds it to get his ideas across. What really ices all this is that neither he nor his students diagnose the problem. They all think that there is an idea here which the students are just having trouble absorbing. If the teacher were to say "Marx's ideas are contradicted by X,Y,Z that you take as economic gospel" then the students would know what knowledge had to be laid aside to appreciate Marx. This is surely better than laying aside either nothing, everything, or a guessed-at selection. Whether teaching Marx, or any other subject, actually works this way I haven't a clue. Educational theorists have a batting average about that of economic theorists, whether you look at the professional leagues or the amateur. -v ------------------------------ Date: 21 Apr 85 08:32:45 EST From: Robert.Thibadeau@CMU-RI-VI Subject: BB Contest [Forwarded from the CMU bboard by Laws@SRI-AI.] Bulletin boards certainly have had their ups and downs. I have always objected to a plethora of boards and to a plethora of posts, a quandary perhaps. But let us invite an "expert system" contest in the world community (for heavens sake!) which provides us with "the intelligent bulletin board". I can indicate my own predilections, and, when I write a message, the system will interpret it with respect to the world of everybody's predilections. Anyone can enter: protocols are at your finger tips. Imagine a system which, after you post your message says, "System won't bother on that post, no one wants to read it." You might even ask why, and find out. Then imagine the thrill, one day, of seeing "System has decided everybody in the world wants that message today -- automatic phoning system engaged." Frabjous joy. This contest is serious: when you have a system design, describe it: consider performance, compatibility, and extensibility along with representational and procedural intelligence. Deadline is December 1, 1985 (mid-term assignments anyone?). Mail description only to prism@cmu-ri-vi. I will put my IRS gift of $100 into a winner and invite other sufferers longing for the good old days to do the same (send commitments not cash). All suggestions will be publicly available, the results published (SIGART Newsletter). ------------------------------ Date: 24 Apr 1985 0959-PST From: GOGUEN@SRI-CSL.ARPA Subject: Seminar - Bertrand Constraint Language (SRI) CSL SEMINAR, 10:30am THURSDAY, 25 APRIL 1985, ROOM EL381 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Bertrand, a General Purpose Constraint Language Wm Leler Computer Research Laboratory Tektronix, Inc. Constraint languages and constraint satisfaction techniques are making the problem solving abilities of the computer available to a wider audience. For example, simple spread-sheet languages such as VisiCalc allow many different financial modeling problems to be solved without resorting to programming. In a conventional language the programmer must specify a step-by-step procedure for the language interpreter to follow. In a constraint language, programming is a descriptive task. The user specifies a set of relationships, called constraints, and it is up to the constraint satisfaction system to satisfy these constraints. Unfortunately, constraint satisfaction systems have been very difficult to build. Bertrand is a general purpose language designed for building constraint satisfaction systems. Constraints are solved using rewrite rules, which are invoked by pattern matching. Bertrand is similar in expressive power to relational languages such as Prolog, but without any procedural semantics. Its lack of procedural semantics makes Bertrand especially attractive for execution on parallel processors. This talk will review several example constraint satisfaction systems built using Bertrand with applications in graphics, design, and modeling. There will also be some discussion of the language issues involved in the design of Bertrand. ------------------------------ Date: Tue 23 Apr 85 13:19:28-PST From: Carol Wright/Susie Barnes <WRIGHT@SUMEX-AIM.ARPA> Subject: Seminar - Assumption-Based Truth Maintenance (SU) SIGLUNCH DATE: Friday, April 26, 1985 LOCATION: Chemistry Gazebo, between Physical & Organic Chemistry TIME: 12:05 SPEAKER: Johan Dekleer Member of Research Staff in Qualitative Physics at Xerox Park TITLE: An Assumption-Based Truth Maintenance System This talk presents a new view of problem solving motivated by a new kind of truth maintenance system. Unlike previous truth maintenance systems which were based on manipulating justifications, this truth maintenance system is, in addition, based on manipulating assumption sets. As a consequence it is possible to work effectively and efficiently with inconsistent information, context switching is free, and most backtracking (and all retraction) is avoided. These capabilities motivate a different kind of problem-solving architecture in which multiple potential solutions are explored simultaneously. This architecture is particularly well-suited for tasks where a reasonable fraction of the potential solutions must be explored. ------------------------------ End of AIList Digest ********************