kp@uts.amdahl.com (Ken Presting) (04/07/90)
(Watch the Followup-To: line) In article <1990Apr6.234803.17071@caen.engin.umich.edu> zarnuk@caen.engin.umich.edu (Paul Steven Mccarthy) writes: >>(Ken Presting) writes: >>I have a (very small) problem with viewing novice+Kasparov as a "system", >> [ ... humans _choose_ to play (and can walk away at will) >> ... the computer _must_ play (laws of physics, no choice) ...](pm) > >You are, of course, assuming that the humans have "free-will". A topic >that I would like to immediately drop now that the assumption has been >identified. OK - you hold it down, and I'll kick it - I'm a major determinist, all the way down to the little hidden variables (bless their hidden little hearts, and the hidden values in 'em). I'll take my non-local lumps back on sci.philosophy.tech, thank you. I mentioned this "very small problem" because there are some who hold a position called "anomalous monism" (I do not) which claims that while all events may be determined by easily statable laws when the events are identified in *physical* terms, while if descriptions in *psychological* terms are supplied to identify the events, there can be no law-like generalization to allow inference from cause to effect. I think anyone who's not up to a few lumps ought to have a small problem with the systems reply. Local random lumps would do, but you'll break the little hearts of the hidden variables - they die if nobody believes in them. :-) ***--- Back to reality ... >As far as whether or not the computer can choose do something else, >consider playing a game of chess on a time-sharing system, . . . I knew I should have mentioned this obvious counterexample, which you are correct to bring up. The chess-playing thread could easily get killed by a kernel short of descriptors, or another user, or swapped out to never-never land ... >dollars to the appropriate level of hardware -- and let the computer >decide what _IT_ thinks is important. (I believe Doug Linnet -- "the >computer that 'discovered' primes" -- is doing just that.) Prioritizing, >choosing goals to pursue / discard is an essential element of >"intelligence", but it seems well within the reach of the symbol >hypothesis. Agreed completely. I'd say that the issue reduces to whether a deterministic account can be given of the conditions under which a process will assume a given trajectory (in state space, phase space, or performance space, whichever is appropriate). > >>Ken Presting ("Support the Symbolese Liberation Army - Out With Symbols") > >Symbol-processing is an insufficient model in my mind as well, but >let's take special care when trying to define its limits. I meant "out" in the sense of "output". I almost said "In & Out With Symbols", but for some reason thought better of it ... Serves me right, not funny enough, again. I see no limitation to symbol processing (except speed) as long as "what TM's do" defines "symbol processing". I just think that "being a symbol" is an *intentional* property, and cannot be applied to any activities of existing AI demo projects. Linnet is close, but MVS is closer. > >---Paul... (I'll take the symbols you throw out -- until I get something > better. "Spare symbols, anyone?" :-) Ken Presting ("Yes, we have no symbols's, we have no symbols's today")