Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!wasatch!cs.utexas.edu!rutgers!att!ihlpb!arm From: arm@ihlpb.ATT.COM (Macalalad) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Message-ID: <9739@ihlpb.ATT.COM> Date: 27 Feb 89 17:32:44 GMT References: <4298@pt.cs.cmu.edu> <4296@cs.Buffalo.EDU> <1989Feb20.213329.10376@cs.rochester.edu> <855@jhunix.HCF.JHU.EDU> <17923@iuvax.cs.indiana.edu> Reply-To: arm@ihlpb.UUCP (55528-Macalalad,A.R.) Organization: AT&T Bell Laboratories - Naperville, Illinois Lines: 83 In article <17923@iuvax.cs.indiana.edu> dave@duckie.cogsci.indiana.edu (David Chalmers) makes the following distinction: >(1) A "symbol" is a formal object which corresponds to some HIGH-LEVEL, >semantic concept in the real world. Typically the concept which it >corresponds to is on the level of a _word_, say, as opposed to a >microstructural level such as that of a neuron. > >versus > >(2) A "symbol" is any formal object which is manipulated by a computer >program. What we take this symbol to correspond to may be as low-level >or as high-level as we like, or we may decide that the question of what >the symbol corresponds to is meaningless and unimportant. I'm not sure if this is a useful distinction. I'll grant that many AI programs deal with symbols at level (1), that perhaps people think only of symbols at level (1) when thinking about the Chinese room scenario, and even that Searle himself might have originally had level (1) in mind when he came up with the Chinese room scenario (although I doubt it). However, this distinction is not crucial to Searle's argument, and I don't think he appeals to our intuitions about symbols at level (1) to argue against symbols at level (2), as you seem to imply: >But here is Searle's trick, or to be charitable (or uncharitable?) his >mistake. He uses the word "symbol" in the low-level sense (2), while >appealing to our intuitions about symbol-manipulators which manipulate >high-level symbols of sense (1)! He says, (paraphrasing), "such a formal >manipulator could never capture the SEMANTICS of the world to which >the symbols correspond." What he implies here is that the symbols >correspond to objects which have meaning, but that formal manipulation >can never capture that meaning. No, what Searle is appealing to is our intuition that there is something more to understanding than just manipulating symbols around. He is arguing that we are something more than just mere formal systems, and that all formal systems, no matter what level their symbols are, lack something else, something essential to understanding. On the other side of the fence, we argue that formal systems are more powerful than what our intuitions lead us to believe. Understanding can "emerge" from a powerful enough formal system. Now we come to a standoff, with each side convinced that the other side is wrong. One of the problems is the Turing Test itself. The Turing Test is essentially a behavioral test which treats the system in question as a black box. The Turing Test judges the system solely on its behavior, without regard to how the system works. Because of this, the Turing Test is vulnerable to skeptical attacks such as Searle's. If we take behavior to be the only criteria for demonstrating understanding, someone can always make the argument that a system isn't really understanding, even if its behavior is very convincing. This eventually boils down to the other minds problem. On the other hand, Searle can't conclude that the Chinese room doesn't understand Chinese; he can only appeal to our intuitions about what can and cannot understand. Obviously, different people have different intuitions. Where do we go from here? There are several options: 1. Ignore Searle's attack and continue building AI systems which come closer and closer to behaving like a human. Unfortunately, unless the system is very, very good, it won't convince anyone that it's understanding, least of all Searle. 2. Acknowledge Searle's attack and build AI "tools" which have no claims to understanding. 3. Sidestep Searle's attack by "strengthening" the formal system. This could be done by adding analog states, sensory input/output, etc. However, if the ultimate criteria for understanding remains a behavioral one, then skeptical attacks like Searle's cannot be avoided. 4. Come to a better understanding of the process of understanding, and different criteria for judging a system than just a purely behavioral one. I think that the most satisfactory answer to Searle would be "This and that are what is involved in understanding, and the formal system in the Chinese room demonstrates (or fails to demonstrate) this and that here and there." I think that most good AI research is done under this category, where the emphasis is on understanding how the mind works, biologically, psychologically, and computationally. -Alex