From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!mcsun!uknet!newcastle.ac.uk!turing!ncmh Tue Nov 19 11:10:43 EST 1991
Article 1380 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca sci.philosophy.tech:982 comp.ai.philosophy:1380
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!mcsun!uknet!newcastle.ac.uk!turing!ncmh
>From: Chris.Holt@newcastle.ac.uk (Chris Holt)
Newsgroups: sci.philosophy.tech,comp.ai.philosophy
Subject: Re: Daniel Dennett (was Re: Commenting on the pos
Message-ID: <1991Nov18.152531.4246@newcastle.ac.uk>
Date: 18 Nov 91 15:25:31 GMT
References: <1991Nov15 .003438.11323@grebyn.com> <1991Nov15.160741.5495@husc3.harvard.edu> <JMC.91Nov17135110@SAIL.Stanford.EDU> <1991Nov17.190935.5546@husc3.harvard.edu>
Sender: news@newcastle.ac.uk (NEWS system)
Organization: University of Newcastle upon Tyne, UK, NE1 7RU
Lines: 43
Nntp-Posting-Host: turing


        [rec.arts.books deleted from Newsgroups line]

zeleny@brauer.harvard.edu (Mikhail Zeleny) writes (to John McCarthy):
>                     ...   I said that any AI researcher stands in need of
>an adequate semantical theory that would characterize the relevant
>relations of expressing and denoting, and could be implemented by a finite
>state automaton, and that, so far, you have failed to come up with an
>answer to this challenge.

Um, what's wrong with the following scenario?  Knowledge acquisition
consists of the registering of primary sensory data by a neural net,
that organizes it and finds regularities.  These regularities can be
thought of as loci in a concept space, with fuzzy boundaries; as more
input is processed, the inherent ambiguity is reduced.  When two loci
are sufficiently distinguishable (the ambiguity between them crosses
an arbitrary threshold), they are understood as distinct, and treated
as points rather than balls (much as very small open intervals on the
real line can be approximated as points).

Symbols are used to refer to these points, and to structures among
them (and for other things); hence denotation.  Structures of symbols
are used to denote these points, and structures among them (and
other things); hence expression (a la data base query languages).
The only difficulty with finite automata is that they limit the ultimate
granularity when concepts are being refined; but since there is only
a finite amount of time per person, and most of that is not used in
organization, this seldom poses a problem.

Consciousness is the property that some of the input to the "mind"
consists of some of the internal workings of the mind (there is a
feedback loop here, as well as the one relating actions to senses).
The formalization of knowledge consists of examining regularities
in actions and the internal workings of the mind, and deriving a
structure that roughly corresponds to it; when any remaining
ambiguity is thrown away, it is a formal model.  Is this the sort
of thing you mean?


-----------------------------------------------------------------------------
 Chris.Holt@newcastle.ac.uk      Computing Lab, U of Newcastle upon Tyne, UK
-----------------------------------------------------------------------------
 "Happiness: a good account, a good compiler, and a good definition."


