COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Dear Dave:
C00007 ENDMK
C⊗;
Dear Dave:

	I just found your memo asking for  comments on the article by
d'Imperio.   This would not be difficult for  me, because I have read
Dreyfus's book, and  I have debated  with Dreyfus  as well as  having
reviewed Lighthill's report for the AI Journal.  I have also read the
Roszak book  to which she? refers.  It is just a question of how much
trouble you think  it is worth.   In my  opinion, unless the  article
expresses a substantial body of opinion in NSA or elsewhere in DoD, I
would rather not devote  too much effort to  it.  The main  reason is
that her  views are  rather eccentric  in rejecting  not only AI  but
"those  often unspoken assumptions  about the brain,  the mind, human
knowledge, and  the nature  of reality  that underlie  not only  A.I.
research, but  in fact our  whole Western scientific  world-view." If
DoD  were  to  follow Roszak  and  give  that up,  it  would  give up
mechanical devices such as rifles and rely on  incantations to defend
the  country.   Moreover,  her summary  of Dreyfus  is  not accurate,
because Dreyfus is not a wholehearted follower of Roszak.  Therefore,
it would be necessary to decide whether  to defend AI against Roszak,
against Dreyfus or d'Imperio.  Just in case you haven't seen it, I am
sending you a copy of my review of the Lighthill Report. 

	Here, however, are some opinions on some of the issues raised
by d'Imperio. 

	1. The success  of AI has been limited so  far.  Comparing it
with genetics, the equivalent of the discovery of the genetic code in
1953 after 100 years of  research hasn't occurred yet.  To  take this
as a reason for  abandoning AI research requires additional arguments
some of which Dreyfus attempted  to supply although his own  position
on whether the research  is worthwhile is rather ambiguous.   I think
that,  on the whole, he  supports the research as  worthwhile for its
potential achievements in areas that he thinks doable, even though he
thinks true intelligence is not achievable by present methods. 

	2.  The issue  of whether  AI research  has enough  payoff to
justify  continued DoD support  is not addressed by  d'Imperio, and I
have addressed it in our proposals and elsewhere. 

	3. Dreyfus's book is not  a very well reasoned attack  on the
ideas  of how  human level  intelligence might  be achieved  that are
current in the AI community.   I could do  it better, i.e. argue  for
the conclusion that  the present ideas are inadequate,  and something
new has to be invented. 

	4.  The  attack  on  AI is  part  of  the  attack on  science
generally that  stems,  it  seems to  me,  from the  fact  that  many
intellectuals  find themselves  unhappy  and  are experimenting  with
blaming  various  aspects  of our  civilization.    AI  is especially
attacked, because it  carries the scientific  method into a new  area
that  remains dear  to  the literary  intellectual,  and because  its
limited success makes it somewhat vulnerable. 

	I can examine the  details of d'Imperio's  points if this  is
important; i.e. if someone important says  that some decision depends
on it. 

				John McCarthy