COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M0BASL30\M1BASI30\M2NGR40\M3NGR25\M4NGR20\MFSTA200\
C00006 00003	
C00007 ENDMK
C⊗;
\\M0BASL30;\M1BASI30;\M2NGR40;\M3NGR25;\M4NGR20;\MFSTA200;\;
\'3;↓↓\FFS\F2
\'3;↓Q\CSTANFORD UNIVERSITY
\F3\CSTANFORD, CALIFORNIA 94305
\F4



ARTIFICIAL INTELLIGENCE LABORATORY\←L\-R\/'7;\+R\→.\→S   Telephone:
COMPUTER SCIENCE DEPARTMENT\←S\→.415-497-4430
\F0\C10 January 1975







Mr. Philip A. Femano
133 Ivy Street
Kearny, New Jersey 07032

Dear Mr. Femano:

\J	Thanks for your thoughtful letter which I only have time to answer
in this cursory way.  Here are my responses to some of the points it raises.

	1. Indeed, will is very important to understanding human behavior.
The human motivational structure is quite complex.  We are descended from
animals that acted over short periods of time to reduce immediate drives of
hunger, fear, sex, etc.  It is amazing that humans can have purposes that
motivate them for years.  I would guess that our primitive ancestors had
some other basic drives besides those mentioned above such as curiosity,
adventurousness, and desire for approval.  Moreover, in humans a secondary
motivation that started in order to help satisfy a primary motivation can
become independent of its source.

	However, it looks like the problem of artificial intelligence can
be separated from this motivational complexity.  We can study intelligence
in connection with systems that have in a single goal the subgoals of which
remain subordinate to it.
This would be a mistake if intelligence depended for its existence on
complexity of motivation.  Moreover, if in humans intelligent behavior
is tightly bound with the motivational structure, an artificial intelligence
that exhibited human-level problem solving ability in a simple motivational
structure might not be very informative about the structure of human
intelligent behavior.

	2. A similar remark can be made about the physiological basis of
intelligence.  Almost all the difficulties now being faced in making
machines solve particular classes of problems appear to have nothing to
do with the physical structure of the machine doing the solving.  Again
the applicability of success in artificial intelligence to human intelligence
may be limited if we ignore physiological considerations.   Nevertheless
I would be surprised if success in artificial intelligence didn't aid
the study of natural intelligence, because it is my conviction that the
methods used by man or machine for dealing with a kind of problem are
determined in large measure by the problem itself.

	3. Artificial intelligence, if and when successful, might not
be "an inferior imitation" of natural intelligence.  There is no reason
to suppose that human intelligence is the most powerful possible
intelligence.  Increasing our understanding of human intelligence is only
one of the motivations for studying artificial intelligence.\.

\←L\→S\←R\-L\/'2;\+L\→L
Sincerely yours,




John McCarthy
Director, Artificial Intelligence Laboratory
Professor of Computer Science
\←S\→L
JMC:pw
femano.le1[let,jmc]:SU-AI