Path: utzoo!attcan!uunet!mcvax!ukc!etive!aiai!jeff From: jeff@aiai.ed.ac.uk (Jeff Dalton) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Message-ID: <223@skye.ed.ac.uk> Date: 28 Feb 89 21:38:40 GMT References: <4298@pt.cs.cmu.edu> <4296@cs.Buffalo.EDU> <4307@cs.Buffalo.EDU> Reply-To: jeff@aiai.UUCP (Jeff Dalton) Organization: AIAI, University of Edinburgh, Scotland Lines: 36 In article <4307@cs.Buffalo.EDU> sher@wolf.UUCP (David Sher) writes: >I'd like to hazzard an answer to this question. The reason the AI >establishment tries to answer this question is there is a strong implication >that Searle's argument indicates that symbolic AI approaches will always >lack some performance capability. >Does anyone believe that they can build a machine with a soul? It is >just as easy to build in Searle's "understanding." It's certainly true that it's hard to see what could ever convince Searle that anything had understanding. But I think we can look at a simpler situation and see the kind of thing that might be involved when there's no performance difference. Let's take Chess. At one time, Chess may have seemed a good test of intelligence. But suppose we have two programs, both able to play at the same level. One program constructs strategies and plans. It explicitly represents goals that, if attained, would trap the enemy kind, and so on. The other just uses "brute force" search, but is very fast. It may behave as if it has goals, but it doesn't really have them (in some sense). Both of these programs are just doing symbol manipulation, and both can play at the same level; but we can still see that they work in different ways. Indeed, the second program is more "mechanical". Both programs can be seen as using only simple, low-level operations; but the first program can also be analyzed in terms of goals and plans while (let us suppose) the second one cannot. That is, the structure of the program doesn't show that kind of organization -- it's behavior is another matter. So I think we can imagine cases where there is no performance difference but where there are other interesting differences we can discover. -- Jeff