kenp@ntpdvp1.UUCP (Ken Presting) (07/11/90)
> In article <593@ntpdvp1.UUCP> kenp@ntpdvp1.UUCP (Ken Presting) writes: > > Searle is trying to prove the following: > > > > For any program P whatsoever, and for any machine M whatsoever, > > the following inference is always invalid: > > > > Machine M runs Program P, therefore Machine M understands. > > (Daryl McCullough) writes: > . . . > . . . the validity of the above inference is not claimed by Strong > AI (or if it is, then they are just speaking loosely). The more > precise claim would be that, for the right program P, one can infer > > Machine M runs Program P, therefore the system (Machine M running > Program P) understands. > Daryl, I take it that your only objection to my reading of Searle is that "System" should replace "Machine" after the "therefore". > . . . For the Chinese room to count as an > argument against this claim, it would be necessary to establish that > the system (man + rules + room) does not understand Chinese. And > Searle cannot establish this without offering *some* definition of > what it means for a system to understand. (Comment: Searle's variant > of having the man memorize the rules does not change anything; there > would still be two systems: the man "acting himself" and the man > following the rules. Establishing that one system does not understand > does not automatically establish that the other doesn't.) > > Daryl McCullough > As I see it, the Systems Reply suffers from a far worse problem of vagueness than does Searle's original argument. What is a "System"? All of us programmers have a very robust notion of how a computer runs a program. We are not confused by such complications as virtual memory, multitasking, distributed processing, and asynchronous execution. (Well, we do get confused sometimes, but whoever it is that wrote the operating system must have figured it out at least once). There is a big difference between using the concept of "System" in the engineering context which originated it, as opposed to using the same concept in an unrelated philosophical context. Here is a simple way to see the problem. You said that when Searle memorizes the rules, there are two systems, one of them being Searle acting as himself. You have also said that a "System" is a machine plus a program. But the question being debated is whether intelligence is constituted by programs. Your version of the systems reply does not *directly* beg the question, but you are making a very big assumption, just by describing Searle-as- himself and Searle-following-rules in the same terms. If you could justify the application of the term "System" to human beings, you would be well on your way to refuting Searle. The problem is identifying just what "program" a natural object like a human brain is "running". You can't take a core dump and disassemble the object code. In _Representation and Reality_, Hilary Putnam presents a "theorem" in which he argues that every physical object instantiates every finite automaton. If he is right, then the Systems Reply has some big troubles. Ken Presting ("Speaking in fork()'ed tongues")
llama@eleazar.dartmouth.edu (Joseph A. Francis) (07/11/90)
In article <598@ntpdvp1.UUCP> kenp@ntpdvp1.UUCP (Ken Presting) writes: >As I see it, the Systems Reply suffers from a far worse problem of >vagueness than does Searle's original argument. What is a "System"? >All of us programmers have a very robust notion of how a computer >runs a program. We are not confused by such complications as virtual >memory, multitasking, distributed processing, and asynchronous execution. >(Well, we do get confused sometimes, but whoever it is that wrote the >operating system must have figured it out at least once). This vagueness only occurs because of our limited understanding. In the CR, our understanding is limited because the decription of the system is so sketchy (we don't really know how the rule book is interacting with the symbols). In human machines (so to speak), it gets even more complex. What portion of you, for instance, understands? It would actually be a more telling blow to the systems response if it did NOT leave us with questions about what it is that constitutes the system. "Read My Lips: No Nude Texans" - George Bush clearing up a misunderstanding
daryl@oravax.UUCP (Steven Daryl McCullough) (07/12/90)
I am a little embarrassed about participating in the to the interminable Chinese Room argument, but I (foolishly) feel that a teeny, tiny bit of progress is being made in the direction of clarifying the issues. In article <598@ntpdvp1.UUCP>, kenp@ntpdvp1.UUCP (Ken Presting) writes: (> > is me, Daryl McCullough) > > . . . For the Chinese room to count as an > > argument against this claim [a version of Strong AI], it would be > > necessary to establish that > > the system (man + rules + room) does not understand Chinese. And > > Searle cannot establish this without offering *some* definition of > > what it means for a system to understand. > > As I see it, the Systems Reply suffers from a far worse problem of > vagueness than does Searle's original argument. What is a "System"? Vague or not, the fact that the Systems Reply is possible at all means that Searle has failed to prove his point. Searle answers the question "Does the Chinese Room understand?" with the seeming non sequitur "The man in the Chinese Room doesn't understand." It seems to me to be up to Searle to show that his answer is relevant; he must show that the understanding of the man is relevant to the understanding of the room, which he cannot do without giving a more complete definition of what it means to understand. He seems to be relying on a terribly naive (since it leads to infinite regress) principle: For an entity to understand, some smaller part of that entity must understand. There is nothing vague or question-begging about denying that the above principle is self-evident, which is all that the Systems Reply amounts to. To defeat Searle, one must only show that his "proof" is not a valid proof. It is not necessary (as you yourself have pointed out) to show that its conclusion is incorrect, only that the reasoning is inconclusive. Therefore it is not necessary to define "system" in order to defeat Searle. > There is a big difference between using the concept of "System" in the > engineering context which originated it, as opposed to using the same > concept in an unrelated philosophical context. The claim of Strong AI, it seems to me (I am just a kibbitzer, not a researcher or philosopher) is that the same notion of system can and should be used in engineering and in philosophical contexts. A defender of Strong AI (which, at least for the moment, I am) would claim that the context is *not* unrelated---that intelligence and understanding *are* essentially software engineering issues. > Your version of the systems reply does not *directly* beg the question, > but you are making a very big assumption, just by describing Searle-as- > himself and Searle-following-rules in the same terms. If you could > justify the application of the term "System" to human beings, you would > be well on your way to refuting Searle. Circular reasoning, or "begging the question" is the fallacy of proving a point by (usually indirectly) assuming the conclusion. I don't think that Strong AI is, in anyone's mind, something that can be proved; it should have the status of a thesis, such as Church's thesis, or a hypothesis, such as Newton's law of universal gravitation. Such statements can't be proved, but one can come up with evidence for or against them. I feel the Systems Reply has the same status. You are right that I can't justify the application of the term "System" to a human being; it is an assumption, or working hypothesis that we can try out to see if it fits, to see if it has interesting consequences, or to see if it has consequences which are clearly false. > The problem is identifying just what "program" a natural object like a > human brain is "running". You can't take a core dump and disassemble the > object code. In _Representation and Reality_, Hilary Putnam presents > a "theorem" in which he argues that every physical object instantiates > every finite automaton. If he is right, then the Systems Reply has some > big troubles. If you temporarily (for the sake of argument) assume that intelligence (and more generally, all mental properties) is a function of software, then an immediate consequence of this assumption is that it doesn't make sense to talk about a physical object (such as a computer, or a human brain) as having one unique intelligence, just as a computer (through multitasking) can run more than one program. To my mind, Putnam's theorem (which I find completely plausible) doesn't defeat the Systems Reply at all; it simply implies that there must be more than one (infinitely many, actually) systems associated with any given physical object. This seems to be exactly what is going on in the thought experiment in which Searle memorizes the Chinese Room rules: Searle's body would then implement simultaneously (1) a staunch opponent of Strong AI, and (2) an artificially intelligent Chinese-speaking mind. To interact with the first system, speak in English, and to interact with the second, write in Chinese. What this comes down to is that a System is not specified by pointing to a physical object; an AI researcher cannot simply hand you a computer and a floppy disk and claim that he has handed you an artificially intelligent being. A specification of a system must also involve a specification of its interface to the world: how it is to be "plugged in", how one talks to it, etc. Daryl McCullough