X-Message-Number: 11529 From: Date: Thu, 8 Apr 1999 11:35:46 EDT Subject: The Chinese Closet Mike Perry writes: >I think Searle makes a fundamental error with the Chinese room experiment, which is to confuse emulator and emulatee. To those on the outside, it appears that the person inside understands Chinese, yet the man inside understands no Chinese, but is just following elaborate rules. The man, in effect becomes the emulator of the real Chinese conversationist. Under appropriate conditions it would, I think, become reasonable to say that (1) there is a conversationist in the room who does understand Chinese, (2) the man is a device that emulates this conversationist but does not understand what the conversationist understands, and thus (3) the man in the room is not the conversationist. Instead of the Chinese Room--a bad metaphor--look instead at the Chinese Closet. This is a very small Chinese Room, with nothing inside except a primitive computer using one simple rule and a modest data store: "Answer string Qn with string An." In a very short conversation, it might answer appropriately, and thus give the impression of intelligence; yet clearly, by any ordinary criterion, it understands nothing. Despite a plausible short conversation, it emulates nothing, not even an idiot. From this point of view, I believe, Searle was right. >As for the more general point that Bob makes, basically, that feeling is not necessarily captured in an emulation, I would again bring up the possibility, which we have no way of disproving, that we are right now in an emulation of some sort, This "possibility" exists only if we concede ahead of time the very point at issue! Further, as I have noted at more length previously, there IS in principle the possibility of determining by experiment whether we are simulations. A simulation cannot anticipate new discoveries in science; the program can only include what was known when the program was written, and deductions therefrom. Therefore there is an insuperable natural barrier for a simulation that does not exist for a "real" entity. This in itself does not preclude the possibility of feeling in computers, but it undercuts the erroneous notion that a simulation from the inside is in principle indistinguishable from "reality." >So I don't accept the argument that feeling is something specifically physical that can't be captured, equivalently, in a different substrate. I think Mike has overlooked my point that real time correlations may be essential to feeling, and these cannot exist in Turing computers. Feeling may depend on time-binding and space-binding constructs. > I do think, once again, that there is no way to resolve this issue by scientific "proof" or "almost-proof" as we usually understand it. And again I point out that, once we understand the anatomy and physiology of feeling in mammals, this will tell us for sure what is sufficient for feeling, and it may well also give us clues as to what is necessary. Robert Ettinger Cryonics Institute Immortalist Society http://www.cryonics.org Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11529