Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!rutgers!elbereth.rutgers.edu!harnad From: harnad@elbereth.rutgers.edu (Stevan Harnad) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Summary: On Turning a Simple Argument into a Complicated Misunderstanding Keywords: Understanding, Comprehension, Learning Message-ID: Date: 3 Mar 89 04:56:37 GMT References: <4298@pt.cs.cmu.edu> <9770@ihlpb.ATT.COM> Organization: Rutgers Univ., New Brunswick, N.J. Lines: 105 arm@ihlpb.ATT.COM (Macalalad) of AT&T Bell Laboratories writes: " (1) In order to understand, a system must be able to introspect. "Introspect"? Let's not unnecessarily multiply our mysteries: In order to understand, a candidate must experience [ = feel, undergo the subjective state of] what we experience [feel, undergo the subjective state of] when we understand. (Prerequisite: It must be able to EXPERIENCE [feel, undergo subjective states] simpliciter.) " (2) A given entity is the best judge of what it can or cannot understand, " given that the entity is capable of introspection. If a candidate is capable of experience at all, it is the only one that can know it. " (3) From (1), in order for the Chinese room to understand, there must " be an introspecting agent. All this fancy formalism and inference is not necessary: For a candidate to understand, someone/something must be experiencing understanding. " (4) The human in the Chinese room is clearly capable of introspection. " (If not, substitute yourself for the human in the Chinese room.) Agreed -- except for the unnecessary, uninformative extra mystery term "introspection": Humans can understand (and in this case, understand English but not Chinese). " (5) From (3) and (4), the human is the introspecting agent in the formal " system, if indeed the formal system has one. This is getting too complicated. What does "having an introspecting agent" mean? Why clutter a simple, straightforward argument with arbitrary, point-obscuring extra baggage? Searle is in the room, doing everything the computer does, but understanding no Chinese. Therefore the computer understands no Chinese (or anything at all) when it's doing the very same thing. " (6) From (2) and (5), the human is the best judge of what the system can " or cannot understand. "The human," in case you've forgotten, is the only one in there, besides the chalk and blackboards! I'll let you be the judge of how good a judge the chalk, or Searle-plus-chalk makes... " (7) The human, upon introspection, concludes that he or she does not " understand Chinese. Have it your way. I'm satisfied to say the only thing in sight (and the only one doing anything) doesn't understand Chinese. " (8) From (6) and (7), the system does not understand Chinese, although " it appears to outside observers that it can. Don't forget that there was a PREMISE in all this, which Searle adopted, from Strong AI, FOR THE SAKE OF ARGUMENT, which was that the LTT (sic) could be successfully passed (till doomsday!) by symbol manipulation alone. Hence you are merely reading back the premise when you remind us of the surprising fact that the Chinese LTT is being passed, i.e., the symbols coming out are consistently and coherently interpretable as discourse from an out-of-sight Chinese interlocutor. It is of course quite possible that this premise is false. (I, for one, believe it is false, and in my paper I give reasons why.) But repeating the premise alone does not invalidate Searle's argument: On the contrary, Searle's argument goes some way toward invalidating the LTT. " Do I fairly characterize Searle's argument? If so, I think that (5) is " clearly the weak point in the chain. And although I've seen Stevan " staunchly defend some of the other points, which I personally don't " have serious problems with, I haven't really seen him address this " point, other than to argue that it is obvious to everyone who hasn't " been brainwashed by a Yale education. You characterized a simple argument in a fairly complicated way, with enough arbitrary extra baggage to obscure its simple point. Reread my comment after (5) above and try to think back to life before Yale... " The systems reply focuses on the weakness of (5), stating that the " introspective agent is not the human, but the formal system. If we " then say that the formal system is incapable of introspection, then " why are we going through the exercise of Searle's argument? Aren't " we assuming that the formal system is incapable of understanding in " order to prove that it's incapable of understanding? Look, do you think an inert book of rules is capable of understanding? If you do, then you'll have no trouble believing that stones, chalk, constellations and tea-leaves are capable of understanding too, and I certainly won't be able to prove you wrong. (Animism or panpsychism -- the belief that anything and everything can have a mind -- is the other side of the other-minds problem.) But if we assume that we need a bit more than that -- say, the ability to pass the LTT [sic] -- then Searle's Argument is there to show us that that's just not good enough, because he can pass the LTT for Chinese without understanding Chinese. And he's all there is to the "system." Ref: Harnad (1989) Minds, Machines and Searle. Journal of Experimental and Theoretical Artificial Intelligence 1: 5 - 25. -- Stevan Harnad INTERNET: harnad@confidence.princeton.edu harnad@princeton.edu srh@flash.bellcore.com harnad@elbereth.rutgers.edu harnad@princeton.uucp BITNET: harnad@pucc.bitnet CSNET: harnad%princeton.edu@relay.cs.net (609)-921-7771