COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002		These are some notes aimed at justifying continued 
C00007 ENDMK
C⊗;
	These are some notes aimed at justifying continued 
DARPA support of research
in artificial intelligence in general and the Stanford Artificial
Project in particular.

1. AI as a scientific topic

	Artificial Intelligence as a scientific subject dates only from
Turing's 1950 paper "Computing Machinery and Intelligence", because
nothing serious could be accomplished until electronic computers became
available.  Proper research computers for AI did not appear until ARPA
support made it possible.  By 1967, we and M.I.T. had suitable facilities
and SRI and Carnegie-Mellon acquired good facilities a bit later.

	The ultimate goal of AI research is to be able to make computer
programs with all the intellectual abilities of humans.  This coupled
with the greater speed, memory, and accuracy ought to make possible
a manyfold increase in our problem solving capability.  It would also
raise serious problems of what to do with this resource.

	At present, no-one is close to achieving a machine with all
the human abilities.  The tasks we have been able to program require
only some of the human intellectual mechanisms for performance at the
level we have achieved.  When a task requires only the mechanisms we
understand, machine performance is much better than human.  Some tasks
like chess and proving mathematical theorems can be done with mechanisms
we understand but only to a limited extent.

	The fundamental research task of artificial intelligence is to
identify, study, and program intellectual mechanisms.

	How far are we from achieving human level intelligence?  In my
opinion, it is not possible to give a definite answer to this question
now.  This is because the main problems that have to be solved have
not been fully identified.  It is clear that some major stroke of
genius is required and possibly more than one.  Perhaps when this
occurs, it will take only a few years to achieve human level intelligence.
Perhaps many successive discoveries have to be made.  Therefore, all
we can do is say that it will take between five years and five hundred
years.

	In insisting on this vague estimate, I don't want to imply that
our understanding of the subject has not advanced significantly or that
it won't continue to advance in the near future.  It has and it will,
and these advances have been and will be reflected in an increased
ability to make computers solve problems.  However, it must be understood
that artificial intelligence like physics or chemistry or mathematics
is a long term research project.