Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!swrinde!tank.news.pipex.net!pipex!peer-news.britain.eu.net!newsfeed.ed.ac.uk!dcs.ed.ac.uk!flugga.dcs.ed.ac.uk!iic
From: iic@dcs.ed.ac.uk (Ian Clarke)
Subject: Re: Does AI belong in neural nets?
X-Nntp-Posting-Host: flugga.dcs.ed.ac.uk
Message-ID: <DnLF5n.G7n.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk>
To: komaromi@ucla.edu (Dan Komaromi)
Sender: iic@flugga.dcs.ed.ac.uk (Ian Clarke)
Organization: Department of Computer Science, University of Edinburgh
X-Newsreader: xrn 8.00
References:  <4gm5c9$phk@saba.info.ucla.edu>
Date: Fri, 1 Mar 1996 14:34:28 GMT
Lines: 57

In article <4gm5c9$phk@saba.info.ucla.edu>, komaromi@ucla.edu (Dan Komaromi) writes:
> Why is this newsgroup not simply called comp.neural-nets?
> 
> The whole field of "artificial intelligence" needs to be re-examined in the 
> context of what we have recently learned about the function of the human 
> brain.  Everything we know about the human brain (long-term potentiation, 
> associative learning, plasticity) contradicts the mainstream assumptions 
> behind these fields.
> 
> I would not include "neural networks" in the artificial intelligence
> category, however.  Since its inception in the 1960's, AI has been
> such an over-hyped and dissapointing field, that we should be careful
> not to expand the label to revolutionary models such as neural nets.
> A neural net is built on opposite philosophical grounds as traditional
> artificial intelligence schemes.
> 
As a student of Artificial Intelligence and Computer Science at the
University of Edinburgh, Scotland, I must disagree with three points you
have made here.  
Firstly, you state that "Everything we know about the human brain (...) 
contradicts the mainstream assumptions behind these fields".  I really 
doubt that you can justify this sweeping statement.  While it may be 
safely concluded that the brain does not work in the same way as 
'classical'-AI programs do, there is still little to suggest that 
intelligence can not be achieved by 'classical'-AI.  I would even venture 
that classical AI programs has come much closer to some human behaviours 
than neural networks will ever manage (logical reasoning etc).   

Secondly, you state that you would prefer not to include neural nets in
the Artificial Intelligence category.  Well you are entitled to your
personal preferences, however as a general assertion, this is also
unacceptable.  I would agree that some distinction between "classical" 
AI and Neural Networks must be made, however there is no basis to claim 
that neural networks warrant a separate field, no more than quantum 
mechanics can be considered to be a separate field to physics.

Thirdly, you claim that "AI has been an over-hyped and disappointing 
field".  For such a disappointing field it is strange that the AI
department here in Edinburgh makes five times more profit from research
than any other department, of _any_ type (disappointing or not), in the
United Kingdom.

You claim that neural networks are built upon opposite philosophical
grounds to traditional AI.  In one sense you are correct.
Traditional AI has a 'top-down' philosophy, and neural networks are
built upon a 'bottom-up' philosophy, however, if you believe that
this is sufficient grounds to consider neural networks to be a
completely different field, then you presumably would say that
writing programs in Prolog is not 'programming' because Prolog is
built upon opposite philosophical grounds to most other programming
languages.  You must see now that this is not a sensible viewpoint.


-- 
|IAN CLARKE        I.Clarke@sms.ed.ac.uk  "..until human voices wake  |
|                  iic@dcs.ed.ac.uk       us, and we drown" - Shelley |
|                  I.Clarke@ed.ac.uk      ianc@aisb.ed.ac.uk          |
