Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!howland.reston.ans.net!EU.net!uknet!newsfeed.ed.ac.uk!hwcee!lundie
From: william@icbl.hw.ac.uk (William Allan)
Subject: Re: Can neural nets think
Message-ID: <DAs3M0.H06@cee.hw.ac.uk>
Sender: news@cee.hw.ac.uk (News Administrator)
Organization: ICBL
X-Newsreader: News Xpress Version 1.0 Beta #3
References: <3sem09$o0v@Owl.nstn.ca> <182254011wnr@tropheus.demon.co.uk>
Date: Mon, 26 Jun 1995 11:55:08 GMT
Lines: 64

In article <182254011wnr@tropheus.demon.co.uk>,
   Steve Wolstenholme <steve@tropheus.demon.co.uk> wrote:
>In article: <3sem09$o0v@Owl.nstn.ca>  Jack Godfrey <JGodfrey@fox.nstn.ca> 
writes:
>> Path:  
>
>> I believe that the subject of this letter is fairly self-explanatory, 
>> however I'd like to supply some possible criteria.
>> 
>> It has already been decided that computers can't think, so it would 
>> require a hardware neural net with some degree of miniturization and 
>> complexity. I'd be interrested to see the responses, especially ones 
>> about neural net minds of some form.
>> 
>Computers compute. Thinkers think. Trouble is we haven't invented any
>thinkers yet. NNs are far away from thinkers - very, very far. Brain size 
>NNs could be made now (they would be rather large) but getting them to 
>think is miles away from just training by example. I would predict that 
>the first thinkers will appear in 2048.
>

I'm not an Expert of any sort in this area but did cover Neural Nets as
part of my Honours Degree so I stress this is merely me dumping my own
thoughts on this matter to into bit form.

The original question was, Can neural nets think?. Well the answer is of
course YES! Our own brains (as far as we know) are Neural Nets so NN's
can think. What the question should have asked is whether Artificial 
Neural Networks can think (ANN's). The answer to this really depends upon
what *thinking* is. We don't know what thinking is and therefore saying
that ANN's don't think seems a bit presumptuous. We have no proof that
they do think, but at the same time we have no proof that they don't.

Thinking is a complex subject and has many levels to consider, as humans,
we think, but we can also think about thinking. A question to ask is
whether animals have this level of thought. Certainly they can think, 
but to what degree? Do they merely think instinctively, or do their 
powers of cognition go to the level of meta-cognition. Are they
concious of themselves as well as their environment, do they think to
themselves, "hmmm I wonder what my master is going to feed me today?".

Getting back to ANN's we should really ask ourselves where on the scale
of thought ANN's lie. Today it is safe to assume ANN's don't think as we
do. They don't even come close to your average goldfish, but this doesn't
mean they don't *think* to some degree. ANN's are at the present time
extermely simplistic, they are subject specific and don't learn in
in the way we do. We are taught and also teach ourselves according to 
our own preferences. ANN's are on the other hand force fed by us and 
don't attempt to optimise and refine their own processes.

In summary (and I apologies for rambling on like this) whether an ANN
actually thinks is a difficult question. It is a question to which we
will all have our own opinions whether they be academic, religious or
spiritual. 

What is the unit by which we measure though? If ANN's don't
*think* at which moment in time and which stage in their complexity
will this change. Where is the threshold between computation and 
cognition? How do we reach that threshold? Does it exist?

These are examples of the type of questions we need to answer but
I fear the answers may be a long way off...

_____________________________________________________________________________
