Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!mcsun!ukc!edcastle!aiai!jeff From: jeff@aiai.ed.ac.uk (Jeff Dalton) Newsgroups: comp.ai Subject: Re: Why the Chinese Room doesn't convince Keywords: Searle, Chinese Room, semantics, Putnam, Lakoff, model theory Message-ID: <2080@skye.ed.ac.uk> Date: 19 Mar 90 23:12:09 GMT Reply-To: jeff@aiai.UUCP (Jeff Dalton) Organization: AIAI, University of Edinburgh, Scotland Lines: 62 In article , kp@uts.amdahl.com (Ken Presting) writes: >In article <1784@skye.ed.ac.uk> jeff@aiai.UUCP (Jeff Dalton) writes: >>Moreover, you can move the line between syntax and semantics. For >>example, when a language is described by BNF (or other context-free >>notations) the syntax doesn't say that the number and type of the >>formal and actual paramaters in procedure calls must agree. So >>that rule has to be part of the semantics. Nonetheless, the rule >>*can* be expressed using a more powerful notation for syntax. >>(Look at the Algol-68 Revised report, for example). >I'd say that this is a case of changing the implementation, or changing >the language, rather than any vagueness in the concepts of syntax and >semantics. Notice that in each case, it's clear that a parameter mismatch >is one or the other: a compile-time error, or a run-time error. Although this is probably an extremely minor point, it should clarify that in both cases the error could be detected at compile-time. The difference is whether it would be detected by the parser or by the compiler's "semantic rules". Whether any part of a compiler should really count as semantics is less clear. It certainly seems that Searle would regard the entire compiler, as well a a formal semantics for the programming language, as "syntax". However, the issues I raised (what has to be syntax and what can be assigned to the semantics, the use of rule schemas, etc) do come up in discussions of whether, for example, English can be described by a context-free grammar. >>However, none of this shows that syntax is sufficient for semantics, >>just that the limits of syntax, or of particular syntactic formalisms, >>aren't always immediately clear. >One very important thing that this whole discussion shows is the >possibility of encoding semantic information with a computer program! >I take it that Jeff is agreeing with Searle's Axiom 3, while cautioning >us that Axiom 3 is not as simple as it may appear. I don't know whether I agree with Searle. His argument seems a little too "neat" to be fully convincing. Still I think there may well be something to it, especially when he doesn't try to encapsulate it in a few simple steps. If you start with meaningless, uninterpreted symbols it's hard to see how they end up referring to what we think we do. For example, there's Hilary Putnam's argument about the "Skolemization of everything" in (I think) "Models and Reality" (see his collected papers Vol 3) and _Reason, Truth, and History_, which claims to show (I hope I have this right) that two people could agree on what sentences were true even if, for example, one of them used "cat" and "cherry" in the usual way while the other used "cat" to mean cherry and "cherry" to mean cat. George Lakoff takes this line of reasoning somewhat further in his _Women, Fire, and Dangerous Things_. And if computers started off with uninterpreted symbols, how would they ever know whether "cat" means cat or cherry or whatever? I think it's right to question whether connecting up some cameras, robot arms, etc would do it, because (as Searle points out) they just produce more meaningless symbols from the computer's point of view (so to speak). I'll have to think about this... -- Jeff