Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!snorkelwacker.mit.edu!ira.uka.de!rusux1!mark From: mark@adler.philosophie.uni-stuttgart.de (Mark Johnson) Newsgroups: comp.ai.philosophy Subject: Re: Searle, Strong AI, and Chinese Rooms Message-ID: Date: 21 Nov 90 11:46:12 GMT References: <1990Nov15.204949.12075@Solbourne.COM> <27320@cs.yale.edu> Sender: zrf80385@rusux1.rus.uni-stuttgart.de Organization: IMS, University of Stuttgart Lines: 35 In-reply-to: mcdermott-drew@cs.yale.edu's message of 16 Nov 90 19:58:34 GMT Drew McDermott makes the interesting claim that in Searle's Chinese Room, we wind up communicating with a "virtual person". This raises all sorts of interesting questions, like "What is a virtual person?", and "How is it that real people like you and me might be able to communicate with them?" Presumably such "virtual people" must be "instantiated" somehow on real entities, and maybe they even need to be "connected" to the "real world" in some way to be "real virtual people"? Maybe all of this sounds crazy to you A.I.'ers --- of course the mind is a computer, what else could it be? (But remember during the last century people thought the brain was like a steam engine, controlled by governors and what not: there is a definite tendancy to view the brain as the most complicated machine around). Actually, what I really wanted to do here is point out the relationship between Searle's Chinese Room argument and some recent issues in the semantics of natural language. There are two basic approaches to N.L. semantics. The first trys to understand N.L.U. solely in terms of symbol processing; if only we can come up with the right representations and algorithms we will be able to explain natural language understanding. The second claims that one must focus on the fact that natural language expressions are *about* something: that when I say "The sun is shining" I'm not just telling you about my internal psychological state, but also about things external to me: the relationship between the sun, clouds, and where I happen to be sitting, etc. These people take the *situatedness* of natural language to be perhaps its most important property. If these people are right, and if intelligence is like language, then its not just the abstract representations and algorithms used by an entity that make it intelligent, but crucially the way in which these representations are grounded in (related to) external reality. Mark Johnson