perm filename PHILOS.2[S87,JMC] blob sn#839525 filedate 1987-05-03 generic text, type C, neo UTF8
COMMENT āŠ—   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	philos.2[s87,jmc]	Philosophy has been of no use to AI
C00005 ENDMK
CāŠ—;
philos.2[s87,jmc]	Philosophy has been of no use to AI

	AI has in principle a closer relation to philosophy than
do other sciences.  This is because a computer program to achieve
goals in the open-ended common sense world must be provided with
notions, however naive, of causality, action, goal, desire, knowledge,
belief and intention.  Philosophers have been studying these notions
for 2,000 years and philosophy has co-existed with AI for thirty.
One would have liked to hope that philosophical analysis would be
helpful in deciding what information involving these notions should
be put into a database of common sense knowledge.  Unfortunately,
we have got no help from philosophers, and I have reluctantly
reached the conclusion that help is unlikely from the present
generation of philosophers --- or their students, so long as they
remain intellectually loyal to their traditions.

	The problem is the determined superficiality of the philosophical
tradition.  For example, one philosopher at Stanford gave a quarter
long seminar on perception in which he puzzled over the distinction
between ``I saw him go'' and ``I saw that he went''.  At the end of
the quarter I asked if his theory of perception made any distinctions
between seeing and hearing, and he proudly answered no.  Theories
as superficial as this are unlikely to be useful in constructing
programs that can see or hear or even the parts of knowledge
bases that need to contain facts about what can be seen or heard
by observers with given opportunities.

	However, while we haven't received aid from philosophers
we have received comfort from some of them --- Daniel Dennett
mainly.  Perhaps he has gone as far as one can with a purely
literary technique, i.e. without mathematical logical formalism.