Path: utzoo!attcan!uunet!mcvax!ukc!strath-cs!glasgow!gilbert
From: gilbert@cs.glasgow.ac.uk (Gilbert Cockton)
Newsgroups: comp.ai
Subject: Bad AI: A Clarification
Message-ID: <1299@crete.cs.glasgow.ac.uk>
Date: 30 May 88 08:34:58 GMT
References: <1242@crete.cs.glasgow.ac.uk>
Reply-To: gilbert@cs.glasgow.ac.uk (Gilbert Cockton)
Organization: Comp Sci, Glasgow Univ, Scotland
Lines: 72

In article <1242@crete.cs.glasgow.ac.uk> gilbert@cs.glasgow.ac.uk (Gilbert Cockton) blurts:
>Mindless application of the computational paradigm to
>     a) problems which have not yielded to stronger methods
>     b) problems which no other paradigm has yet provided any understanding of.
This is poorly expressed and misleading.  Between "problems" and "which" insert
"concerning human existence".  As this stands, it looks like I want to withdraw
encouragement from ALL computer research.  Apologies to anyone who's taken this
seriously enough to follow-up, or was just annoyed (but you shouldn't be anyway)

Bad AI is research into human behaviour and reasoning, usually conducted by
mathematicians or computer scientists who are as well-qualified for the study
of humanity as is an archaeologist with a luminous watch for the study of
radiation (of course I understand radiation, I've got a luminous watch, 
haven't I? ;-))

AI research seems to fall into two groups:
	a) machine intelligence;
	b) simulation of human behaviour.
No problem with a), apart from the use of the now vacuous term "intelligence",
which psychometricians have failed miserably to pin down.  No problem with b)
if the researcher has a command of the study of humanity, hence the
respectability of computational modelling in psychology.  Also, mathematicians
and computer scientists have no handicaps, and many advantages when the human 
behaviour in b) is equation solving, symbolic mathematics, theorem proving and
configuring VAXES.  They are domain experts here.  Problems only arise when they
confuse their excellent and most ingenious programs with human reasoning.

   1) because maths and logic has little to do with normal everyday reasoning
      (i.e. most reasoning is not consciously mathematical, symbolic, 
      denotational, driven by inference rules).  Maths procedures are not
      equivalent to any human reasoning.  There is an overlap, but it's small
   
2)    because they have no training in the difficulties involved in studying
      human behaviour, unlike professional psychologists, sociologists,
      political scientists and economists.  At best, they are informed amateurs,
      and it is sad that their research is funded when research in established
      disciplines is not.  Explaining this political phenomena requires a simple
      appeal to the hype of "pure" AI and the gullibility of its sponsors, as
      well as to the honesty of established disciplines who know that coming to
      understand ourselves is difficult, fraught with methodological problems. 
      Hence the appeal of the boy scout enthusiasm of the LISP hacker.

So, the reason for not encouraging AI is twofold.  Firstly, any research which
does not address human reasoning directly is either pure computer science, or 
a domain application of computing. There is no need for a separate body of
research called AI (or cybernetics for that matter).  There are just
computational techniques.  Full stop.  It would be nice if they followed
good software engineering practices and structured development methods as
well.  Secondly, where research does address human reasoning directly, it
should be under the watchful eye of competent disciplines.  Neither mathematics
or computer science are competent disciplines.  Supporting "pure" AI research
by logic or LISP hackers makes as much sense as putting a group of historians,
anthropologists and linguists in charge of a fusion experiment.  The word is
"skill".  Research requires skill.  Research into humanity requires special
skills.  Computer scientists and mathematicians are not taught these skills.

When hardware was expensive, it made sense to concentrate research using
computational approaches to our behaviour. The result was AI jounals,
AI conferences, and a cosy AI community insulated from the intellectual
demands of the real human disciplines.  I hope, with MacLisp and all
the other cheap AI environments, that control of the computational
paradigm is lost by the technical experts and passes to those who
understand what it is to study ourselves.  AI will disappear, but the
work won't.  Indeed it will get better, and having to submit to an AI
conference rather than a psychology or related conference (for research
into ourselves), or a computing or application area conference (for
machine 'intelligence') will be a true reflection of the quality of the work.
-- 
Gilbert Cockton, Department of Computing Science,  The University, Glasgow
	gilbert@uk.ac.glasgow.cs !ukc!glasgow!gilbert

	     The proper object of the study of humanity is humans, not machines