Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!husc6!bloom-beacon!apple!rutgers!elbereth.rutgers.edu!harnad From: harnad@elbereth.rutgers.edu (Stevan Harnad) Newsgroups: comp.ai Subject: Re: Question on Chinese Room Argument Summary: On Achilles and the Tortoise Message-ID: Date: 21 Feb 89 02:17:37 GMT References: <4298@pt.cs.cmu.edu> <51157@yale-celray.yale.UUCP> Organization: Rutgers Univ., New Brunswick, N.J. Lines: 123 The careful reader will find an uncanny resemblance between the logic underlying the following exchange and Lewis Carrol on Achilles and the Tortoise (in which Carroll showed that you can lead someone to logical water, but there's no way to make him drink it). Read on: engelson@cs.yale.edu (Sean Engelson) of Computer Science, Yale University, New Haven, CT 06520-2158 asks: " What is your criterion for determining which is which [understanding " or not understanding Chinese]? As I said before, you need no definitions, no criteria. You only need to be able to tell the difference -- in your own, subjective, first-person case -- between when you understand a language (e.g. English) and when you do not (e.g., Chinese). Can you do that? Now please assume that Searle can do the same, and that he says he does NOT understand Chinese. There is no reason whatever (apart from the preconceptions that Searle's Argument was formulated to invalidate) (a) not to believe him or (b) to believe that there is "someone/something" else in the Chinese Room that IS understanding Chinese in the same sense that you or I or Searle understand English. To pick (b) merely on the basis of the preconceptions that are the very ones under criticism here is simply CIRCULAR. -- Now none of this is new; one would have thought that it would be clearly understood (sic) from my prior posting. Read on. " And I can equally well POINT to Searle running his rules for Chinese, " and say to (him + rules) in Chinese, "Look, you understand Chinese, " don't you?" and I'd expect to get back the answer (in Chinese) "Yes". In saying you could point to it, I was clearly speaking about the subjective phenomenon (i.e., whether YOU YOURSELF understand English or Chinese), which is primary and not open to doubt, rather than to its external manifestations (i.e., whether SOMEONE ELSE does): The evidential status of those external manifestations is precisely what's on trial here; one can't win this case by simply declaring them "judge and jury" instead! That's not a logical supporting argument; that's just circularity. " for all external intents and purposes, (Searle + rules) understands " Chinese... [whereas] "plain" Searle does not understand Chinese, " (Searle + rules) does not. Why not? What's the difference? The difference is that the "external" criteria have not been shown to be valid, and hence there is simply no justification for taking them to signal the presence of understanding at all. To merely assume that they do is not an argument; its just circularity again. For all external purposes, we have a (hypothetical, perhaps even impossible) situation in which a guy (imagine it's you) is running around manipulating symbols and saying he can't speak Chinese and has no idea what the symbols mean; meanwhile, ex hypothesi, if a Chinese person reads the symbols, they say "I understand Chinese... etc." Even if we accept the unlikely hypothesis that this is possible (and could go on for a lifetime, with the symbols as consistently lifelike and convincing as a real-life Chinese pen-pal), there's still no one around about whom we could say, "Ya, well if he says he understands, I'm ready to believe he understands, just I am about anyone else who says he understands." The only one around is you, and you say (don't you?) that you don't understand. Perhaps we should ask the hypothetical Chinese alter ego to say where he is, and where he stands on the matter... Part of the problem is of course with the premise itself (i.e., supposing that we could do all this with just symbols), which may be about as realistic as supposing that we could trisect an angle with just compass and straight-edge. All that the premise seems to do is to spuriously mobilize our instincts to defend the personhood of our unseen pen-pals. But recalling that there's no way we can be sure about our pen-pals under such conditions either, and that the SOLE case of understanding we can be sure about is still our very own (i.e., the "other minds problem"), ought to be a good antidote for mistaking external signs for the real thing even under such counterfactual conditions. " In other words, you are DEFINING understanding to be that which Searle " has with respect to English, and not that which (Searle + rules) has " with respect to Chinese. OK, given that distinction, tell me either " how I can distinguish between the two in a Turing-test fashion, or " what it is about Searle that allows him to understand English which " (Searle + rules) does not have. Otherwise, as I've said, I'll grant " you your point, and then say that this whole discussion is pointless, " as your concept of understanding is "That which people do", which is " useless. Have we made any progress here? I think not. I keep saying I'm not defining but pointing to a subjective experience that all people have and Engelson keeps talking about fanciful things that "people plus rules" have. Now he says that all this logical, methodological and empirical discussion, which was originally intended to assess the evidential status of the (teletype version) of the Turing test is now answerable to that test A PRIORI! That's like saying: "Well if God didn't create the earth, then tell me how he created Darwinian Evolution?" Preconceptions manage to survive without ever becoming negotiable! The only other possibility Engelson seems ready to imagine is a complete alternative causal/functional explanation of understanding that distinguishes Searle from a mere symbol-manipulator; but I've already said that we're far from such an account, nor do we need one for present purposes. Logically speaking, you just have to show that a theory is internally inconsistent or inconsistent with the data in order to show it's wrong (although Kuhn will of course remind you that that's not enough to make people give it up). You don't have to come up with the right theory. (If you want a better candidate in this case, though, try grounded hybrid robotic systems, as I've suggested.) Searle's denial that he understands Chinese (or your own denial, if you were in his place and had not yet been at Yale-CS too long to be able to call a spade a spade) seems like a big enough inconsistency to do in the purely symbolic theory. My "concept of understanding" is no different from what yours was before you bought into a fantasy that "a person plus rules" could understand even if the person couldn't. And if you want an idea of just how pointless a discussion is when logical arguments are unavailing, read Lewis Carroll on Achilles and the Tortoise. But to go on like this is more like Schultz on Charlie Brown and the football... -- 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