Path: utzoo!attcan!uunet!husc6!bbn!rochester!uhura.cc.rochester.edu!sunybcs!bingvaxu!leah!itsgw!imagine!rpics!wardeng
From: wardeng@rpics (Greg Warden)
Newsgroups: comp.ai
Subject: Re: Intelligence / Consciousness Test for Machines (Neural-Nets)???
Keywords: Machine Intelligence
Message-ID: <1363@imagine.PAWL.RPI.EDU>
Date: 7 Oct 88 14:25:59 GMT
References: <1141@usfvax2.EDU> <40680@linus.UUCP> <263@balder.esosun.UUCP>
Sender: news@imagine.PAWL.RPI.EDU
Reply-To: wardeng@turing.cs.rpi.edu (Greg Warden)
Organization: RPI CS Dept.
Lines: 19

When can machine be considered conscious?

Of course this is really asking what is consciousness? This may 
have to do with an ability to extract semantic meaning from symbols 
(or stimulus realy) that do not have any inherent meaning already.

You can teach a machine to manipulate english all you want, but when does
it start to KNOW what the characters MEAN? Now this is atleast the 
traditional attack of strong AI (see John Searl _Minds Brains & Programs_).
I am not convinced that this is a valid argument. First if we are to 
consider ourselves conscious, we assumes that we have some cosmic 
understaning about our stimuli. Do we? I am not an expert on theory
of language but how do we get info out of a word? It seems to me that we
refer the word to a referent--sort of a level of abstraction. Maybe
if the machine had data about what words refered to to (and what they
refered to) and it knew what the relationshinp between the words and
referents it would be conscious. 

greg warden