CSE 4/510 & PHI 498, Spring 2004

Position Paper #5: Can Computers Think?

Last Update: 25 March 2004

Note: NEW or UPDATED material is highlighted

For this position paper, I would like you to evaluate the following hypothetical debate.

Pro: If something behaves in all relevant ways as if it were cognitive, then it is cognitive.

Con: What do you mean by "being cognitive"?

Pro: I mean that it can perceive (see, hear, etc.); has beliefs, desires, and intentions; can remember; can use and understand natural language; can reason and make rational decisions; etc. You know, the sort of thing that AI researchers are trying to achieve by computational means.

Con: Do you think they will succeed?

Pro: I'm optimistic: I think that a suitable AI program (or maybe a suite of programs) will eventually behave in all these ways.

Con: But that means that you think that such an AI program will be cognitive?

Pro: Yes.

Con: But that's crazy! Computer programs are purely syntactic!

Pro: Now it's my turn to ask for clarification: What do you mean by "syntactic"?

Con:: I mean that all it can do is to manipulate the symbols of a formal symbol system.

Pro:: So what's the problem?

Con:: The problem is that cognition is semantic! That is, it involves the semantic interpretation of those symbols.

Pro:: Well, I'm not so sure about that. But suppose you're right. What then?

Con: Well, syntax does not suffice for semantics. So, no purely syntactic computer program can exhibit semantic cognition, even if it behaves in all relevant ways as if it were cognitive.

Copyright © 2004 by William J. Rapaport (rapaport@cse.buffalo.edu)
file: 510/pospaper5-2004-03-25.html