Standards in Artificial Intelligence

Nick Maclaren nmm1 at cus.cam.ac.uk
Thu Nov 2 10:41:32 EST 2000


In article <qa66m68m4i.fsf at tardis.ed.ac.uk>, James Hammerton <james at tardis.ed.ac.uk> writes:
|> nmm1 at cus.cam.ac.uk (Nick Maclaren) writes:
|> 
|> Hmmm. I always understood "Artificial Intelligence" to mean the building
|> of intelligent machines.

Well, it SHOULD do - but I have very rarely seen it used in that way.
There is very little current work that can honestly be called the
building of intelligent machines, after all, and a hell of a lot
is called "AI".

|> > the term "AI" was coined by some clueless (if eminent) academic,
|> > hijacked by another of the same class, invented by the press, or
|> > what.  
|> > But it is and always has been nonsense.
|> 
|> Why do you say that?

See above :-)  It is nonsense, because it is used as a sexy term
for things that have nothing special to do with intelligence,
artificial or otherwise.  It is the term that I regard as nonsense,
not (most of) the research.

|> > It was originally a term for genuine research into the limits of
|> > what could be done with computers, often including natural language
|> > processing.  This was and is all good, sound research.
|> 
|> >  It hasn't got very far, but we now have a deeper understanding of the
|> > problems :-)  That class of "AI" researcher usually loathed
|> > being associated with the term "AI".
|> 
|> Not in my experience. 

Definitely in my experience.  But please note that I am talking
about the 1960s and 1970s.

|> > It was then hijacked by some people to use as a sexy name for
|> > some empirical search and optimisation techniques (so-called
|> > neural nets and all that).  
|> 
|> But many of these researchers are tackling the same problems that the
|> researchers you refer to above are tackling, e.g. reasoning, natural
|> language processing, robotics, cognitive modelling. You may disagree
|> with their techniques, but that's another matter.

No problem.  But why should that work be described as artificial
"intelligence" any more than (say) pattern recognition and image
processing?  Yes, we could call all algorithmic work "artificial
intelligence", but that would be silly ....

|> > They predated the term "AI" by many
|> > years, at least as it was used applied to them.  Use of the term
|> > "AI" for this is simply nonsense.
|> 
|> Currently my research involves using neural networks for machine
|> learning of natural language processing tasks.
|> 
|> Why do you think that is fundamentally any different from
|> people using inductive logic programming or probablistic grammars for
|> the same tasks?

I don't.  Why should I?  The term "AI" is as misleading for one
as for the rest.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QG, England.
Email:  nmm1 at cam.ac.uk
Tel.:  +44 1223 334761    Fax:  +44 1223 334679






More information about the Neur-sci mailing list