walt_sak@proper.UUCP (Walt Sakai) (03/09/85)
Prefixed by a ">" below are the responses to discussion #001: (Everyone is encouraged to participate. See "*" note at EOT.) > From: amdahl!amd!twg!scc!steiny (Don Steiny) > I don't think Winogand has much to do with Wittgenstein. His > whole thing about reference, i.e. a block world, is exactly the > opposite of what Wittgenstein proposed. In linguistics courses > the pragmatic aspect of language was how it is used, the semantic > is what it refers to. Don, I assume you are talking about Terry Winograd. I certainly agree that Wittgenstein had little to do with the "block world" model per se. Winograd is taking a methodological approach along the lines of Wittgenstein's philosophical interests. By methodological I mean that the problem is approached in a way that the component parts are examined first. In this case, the block world is taken to be a subset of the natural world. I think Winograd would be the first to admit the limitations of the block world. However, it has the virtue of shedding light on a easier problem than that encountered in the natural world. Scientifically, we would like to first develop models of some finite aspect of a phenomenon, and hopefully, by having a valid model for every finite aspect, we can confidently generalize the model for the phenomenon as a whole. Winograd's block world serves only as a first step towards the formalization and explication of the multitude of factors that constitute context or use (in the sense of Wittgenstein's INVESTIGATIONS) with respect to language. In addition, it seems very evident that both Wittgenstein and Winograd are interested in analyzing the nature of linguistic competence and how that competence is related to the formal rule structures in languages. > From: amdahl!amd!twg!scc!steiny (Don Steiny) -continued > The early Wittgenstein and the positivists like Russell and > Carnap proposed that since words like "justice" do not refer to > anything we could stop worrying about them. > > The question of reference is a central issue in both computers > and linguistics. To maintain the idea that words have reference > people have often invented the object that the words refer to. > The whole idea of "brain states", i.e. "pain is a state of the > brain" has been explored by Hiliary Putnam and Jerry Fodor. > Putnam has a book, I believe (it is out of print and I have to > check it out of the library, I don't have it right now), Mind, > Models, and Machines. In the book there is a lengthy summation of > Wittgenstein's argument against such reference. > > There is a book called "Language and Learning" which is a round > table discussion between Putnam, Fodor, Gregory Bateson, Seymore > Papert, Noam Chomsky, and Jean Piget about their ideas. I think > what falls out of it is the idea that AI and natural language > work on the computer are not the same. Natural language > processing is not a subset of AI, it is a different field. > > Since I write programs that do natural language processing the > issue is of great interest to me. Winogand's representation of > knowledge, semantic nets, and other representations of human > knowledged are based on the premise that words refer. How can we > develop a principled way of writing programs and organizing > knowledge so that the words *signal*? I would be rather inclined to say that some very crucial problems of AI (Artificial Intelligence) are subsets of natural language issues. In natural languages our utterances are made and understood with a tacit knowledge of the world. The speaker and listener share a common background with respect to the workings of the "world". This could include the regularity of certain events (some subsumed under scientific laws) as well as social customs. Misunderstandings (and jokes) can occur when the implied background assumptions are not shared. One of the problems in the interface between AI and natural languages is the formalization and explication of our "tacit" knowledge. It is vast and practically limitless, hence any attempt to codify such knowledge can be expected to fall short of our expectations. (It might help if someone found a recursive structure. :-) A great deal of our tacit knowledge of the world consists of litte tidbits and fleeting impressions, which of course do not fall under principles or any methods of organization. The ultimate test of these general considerations is in the field of computer translation. Essentially the translator must have an understanding of the "background" inherent in two different natural languages. Where the translator is a computer such an understanding must be explicitly codified for the performance of the task. And this is where the great abyss exists. (When we have a successful computer translation of Proust into English, we'll invite the engineer to explain how s/he substituted bio- chemistry for electronics in the microprocessor.) > From nsc!seismo!philabs!micomvax!othervax!psal > Please add me to the mailing list. Anyone out there interested in > Langer or Clauswitz as well? > -C.Thomas Weinbaum von Waldenthal Perhaps you tell us more about Clauswitz. If his interests overlap with those of Wittgenstein's this is the place to introduce him. > From nsc!seismo!mcvax!vu44!botter!botter!biep > Here in Europe we don't get net.philosophy. However I've never > read anything of Wittgenstein, I thing I would be very interested > to follow this discussion, that is: reading in. Would it be > possible for me to receive the articles by mail? Thank a lot in > advance! Biep. > Some mazes (especially small ones) have no solutions. OK, so what net category close to net.philosophy do you get in Europe? There is no mailing list yet. See "*" note at EOT. > From nsc!seismo!mcnc!ecsvax!unbent Wed Feb 20 14:04:51 1985 > Original-From: Jay F. Rosenberg <unbent@ecsvax> > I'm afraid that my Wittgenstein essays all predate my > computerization, so none of them exist in any "electronic" form. > Sorry. I'm afraid that one is condemned to actually hauling them > out of the library in order to read them. That probably explains > why so few people have. > > One of our fellow conferees found Heidegger difficult to read in > English. You may assure him, on my behalf, that Heidegger is just > as difficult to read in German. The best handle on Heidegger that > I know of is HEIDEGGER AND THE PROBLEM OF KNOWLEDGE, Charles B. > Guignon, Hackett Publishing Co. (Indianapolis; 1983). There are > indeed affinities between Heidegger and Wittgenstein, but, as far > as I know, no mutual *influence* in either direction (i.e., no > evidence that either ever read anything by the other). I recall reading something in Wittgenstein's letters about Heidegger. Nothing of great significance though. I imagine Wittgenstein had some idea of Heidegger's BEING AND TIME since portions of it was shown to be nonsense by Carnap in the Vienna positivist days. There has been a MISPRINT in discussion #001, Jay's essays: 1> "Wittgenstein's Theory of Language as Picture", AMERICAN 1> PHILOSOPHICAL QUARTERLY, V,1, January 1968, pp. 18-30. 1> 1> "The Concept of Linguistic Correctness", PHILOSOPHICAL STUDIES, 1> 30, 1977, pp. 171-84. 1> 1> "Speaking Lions", CANADIAN JOURNAL OF PHILOSOPHY, VII,1, 1977, 1> pp. 155-60. 1> The second article should be dated 1976, not 1977. I found the third article by Jay quite lively in contrast to the very dry writings usually found in academic journals. For those not familiar with Wittgenstein, I recommend reading the third article instead of the second. Interestingly, on page 183 of the second article, Jay says: "The existence of persons as rational beings is thus conceptually inextricable from their existence as social beings." And the reason is compactly stated on page 182: "If, apart from his membership in a linguistic community, an individual cannot correctly or incorrectly call some confronted item a 'T', neither, in the absence of such a shared form of life, can he correctly or incorrectly think it to be, believe it to be, judge it to be, or know it to be a 'T'." This ties in with the AI discussion above in the sense that the acquisition of some our tacit knowledge of the world would require a person/entity to be a member of a linguistic community. Question: in what sense are AI systems rational? If so, in what way could they ever be affirmed to be a member of a linguistic community? In the third article, Jay first considers the proposition: "If a lion could speak, we could not understand him." It is found to be a little incorrect, in that we could deduce and confirm what the lion is saying from his reports of external events. The difficulty comes when internal events are reported, eg. feelings and thoughts. Wittgenstein's private language arguments are relevant here. So the proposition is modified to read: "If a lion could speak of himself, we could not understand him." But Jay adds the caveat: "...unless of course we live with this lion, come to know him as a friend, enemy, etc... but then he will be one of us." > From: dual!ihnp4!gargoyle!sphinx:sphinx!thos > (Tom Barron,merr 376,28850,4935765,isdv) > I'm very interested in participating. I spent more years of my > life trying to unravel Wittgenstein than I'd care to admit. > There's a shelf full of paper somewhere on the subject but I'm > going to ignore it, to start fresh, if that's possible. > > I've read (and will reread) Jay Rosenberg's "Wittgenstein's > Theory of Language as a Picture". I benefited from it before and > think it might well provide a way of focussing discussion. > Incidentally, there's a little book Rosenberg wrote called "The > Practice of Philosophy" (I think) that used to be published by > Prentice Hall that I'd recommend to any of our participants who > want to brush up on philosophical techniques. Perhaps I should > conclude with the usual product disclaimers: I've never met > Rosenberg, don't own stock in the University of North Carolina, > etc. Thank you, Tom, for responding. Welcome on behalf of the group. I think we all would be very interested in your thoughts after you re-read Jay's essay. * Discussion Group: LUDWIG WITTGENSTEIN --> net.philosophy * * ---------------------------------------------------------- * * The topics will involve overlapping fields such as linguistics, * * philosophy, and mathematics, in order to maximize the number of * * participants. However, all discussions will be located in * * "net.philosophy", which will serve as a common meeting ground. * * * * I imagine that some communication will be made via "mail" between * * private individuals meeting through the main line of discussion. * * Hopefully, these private discussions will be posted to * * net.philosophy for source of further thought by everyone else. * Walt Sakai {ucbvax,hplabs,ihnp4,cbosgd, decwrl,unisoft,fortune,sun,nsc}!dual!proper!walt_sak
jc@mit-athena.UUCP (John Chambers) (03/18/85)
> In the third article, Jay first considers the proposition: "If a > lion could speak, we could not understand him." It is found to > be a little incorrect, in that we could deduce and confirm what > the lion is saying from his reports of external events. There was a clever illustration of why this is fallacious, published some years ago in the form of a science fiction short story. Sorry, but I don't remember either the title or author; maybe there are some SF fans around who do. The scenario was the discovery of ruins of a recently-dead civilization on Mars, complete with extensive libraries. The only trouble was, there wasn't any Rosetta stone. The libraries had been written by members of a species with no contact whatsoever with humans, not even common ancestors that might have produced minimally similar outlooks on life. But the scientists broke the "code" anyway. The breakthrough came when one of the scientists recognized the layout of a large chart on a wall in a room. It was a periodic table. This is not a function of an individual's outlook on the universe. Hydrogen has the same properties everywhere, regardless of who is observing it. So does helium, lithium, ..., uranium. The electron orbitals are the same everywhere, and they are what determine the observable properties of the elements inside chemical compounds. Any scientists on any planet will come up with a periodical table, and a chart will look the same (except for rotations and reflections). The basic properties (atomic number, atomic weight, filled/unfilled outer orbitals) will be the same numbers (in some representation, to about 10 bits precision). Once such an artifact is encountered, decoding the numbering system would be straightforward; any cryptanalyst could do it. (No, I don't believe there could be a usable numbering system that humans couldn't decode. That's silly.) Most of the names for the elements would be complex symbols with visible inter-relationships, which gives you some good hooks on the low-level syntax of the language (or at least of the writing system). If a lion were intelligent enough to "speak" (i.e., produce noises that communicate information), and were willing to speak to us, there would be relatively little trouble establishing understanding on a wide range of topics. Of course, we might never come to terms on value-judgements. But that's a change-of-subject. -- John Chambers [...!decvax!mit-athena] If you're not part of the solution, then you're part of the precipitate.