home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.philosophy
- Path: sparky!uunet!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!spool.mu.edu!umn.edu!csus.edu!netcom.com!doug
- From: doug@netcom.com (Doug Merritt)
- Subject: Re: The Humongous Look-up Table revisited.
- Message-ID: <1993Jan1.223614.26337@netcom.com>
- Organization: Netcom Online Communications Services (408-241-9760 login: guest)
- References: <1992Dec22.211814.5623@u.washington.edu>
- Date: Fri, 1 Jan 1993 22:36:14 GMT
- Lines: 43
-
- In article <1992Dec22.211814.5623@u.washington.edu> forbis@stein.u.washington.edu (Gary Forbis) writes:
- >I hope there is no question in anyone's mind but that there exists
- >HLTs that will pass the Turing Test. [...]
- >Interogator: What time is it in Seattle?
- >HLT: Oh, about noon. Why?
-
- Several other people answered aspects of this definitively, but perhaps
- not in a 100% clear way, so I thought I'd clarify...
-
- The current time, info about weather, and other such state information
- (call them sensory inputs) *necessarily* will be inputs to any mechanism
- that hopes to pass the Turing Test.
-
- In the case of the lookup table, the fields that represent this current
- state of the world are merely appended to the input sentence, and the
- whole conglomeration is used to index the lookup table. There are no
- theoretical problems, no stepping away from the model of a lookup table,
- no need to enhance the lookup table by allowing it to do field substitutions.
-
- This approach makes the table bigger than fancier mechanisms, but that's
- beside the point. It works in principle. A large enough lookup table can
- compute anything that a finite Turing machine can compute...lookup tables
- are Turing complete.
-
- Nor is it an issue (mathematically) if the input sentences are allowed to be
- of infinite length, nor if the world is continuous, therefore requiring input
- state of infinite precision and therefore infinite length...One just makes
- the lookup table of infinite length as well, and define a one to one mapping
- between the set of inputs and the set of outputs. Mathematically this is
- nearly as trivial as a finite map (lookup table).
-
- Naturally it is dubious that an infinite lookup table could be constructed
- in the physical world, but then it's dubious that a lookup table
- with finite but humongous (e.g. 10^100000000) entries is constructible
- either.
-
- Whether the construction of such tables (finite or infinite) is computable
- even in principle amounts to the question of whether strong AI is possible,
- so there's no point in getting into that here.
- Doug
- --
- Doug Merritt doug@netcom.com
- Professional Wild-eyed Visionary Member, Crusaders for a Better Tomorrow
-