wolit (11/24/82)
(In response to jim@sri-unix's posting of John Searls' "refutiation" of the Turing test) All that this "thought experiment" shows is that the Turing test doesn't localize the intelligence in a system. The intelligence in this case is obviously contained in the book of instructions (on how to combine Chinese characters), not in the human subject. No one would claim that the fingers of a human taking such a test were "intelligent", even though it is they that would work the keyboard, for example, nor that the individual neurons that make up an intelligent brain are "intelligent" (they are, in fact, quite mechanical in their responses). Similarly, there is nothing that says that an intelligent computer could not be composed of millions of very dumb transistors. Douglas Hofstadter (sp?) treats this subject in his book, "Goedel, Escher, Bach"; I'm satisfied that the Turing test can withstand at least this minor challenge. Jan Wolitzky, BTL MH, rabbit!wolit
cbostrum (11/27/82)
The author of the article mentioned in the title of this response is really referring to John R. Searle of UCB and his infamous "Chinese Room" argument against what he calls "strong AI". The paper where he presents this argument can be found in Mind Design, edited by John Haugeland; its title is "Minds, Brains, and Programs". It can also be found in a recent issue of the Behavioural and Brain Sciences Journal, along with commentaries from many others in the field. Many of these commentaries are absolutely devastating to Searle's argument, which, frankly, is astonishing naive for a philosopher of his repute. If I recall, Haugeland's response itself was a particularly good one. One final place to look for this paper is The Minds I, ed. Daniel Dennett et. al. There is some commentary there as well. The basic flaw in the argument is of course that it is **not** the man who is manipulating the symbols to whom the comprehension of Chinese is being attributed, but rather the entire system itself. It is roughly the same mistake as saying that our brain (or some other suitably chosen part of us) understands english, rather than we ourselves. Haugeland compares the man running around pushing symbols on little pieces of paper and eventually dropping one such piece out a slot to the nueral pathways in our brains passing "messages" about; and surely we do not wish to say that these pathways understand english. But this is essentially what Searle is trying to get us to believe. Although I dont believe Searle's argument serves in any way as a refutation of the Turing test, I do believe that the Turing test is not what its cracked up to be. Comments and discussion on this issue would be quite interesting, I think.