COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	hofsta[s86,jmc]		Comments on Hofstadter
C00005 ENDMK
C⊗;
hofsta[s86,jmc]		Comments on Hofstadter
vijay@ernie.berkeley.edu
Comments on Hofstadter

Hofstadter discusses three characterizations of AI but unfortunately
omits the one I and many others favor.  AI is the study of how
problems can be solved and goals can be achieved in common sense
environments.  In this it is like analysis of algorithms in being
conceptually a mathematical subject.  Unfortunately, we still have
no full characterization of the kinds of environment in which
AI programs must operate and in which humans achieve their impressive
performance.  Nevertheless, for AI we don't need to fully imitate
human mental qualities or even to fully understand them.  It's
just that, for the present, contemplating natural intelligence
is useful in developing artificial intelligence.

I agree with Hofstadter that there is a lot still to be learned
from the blocks world, but I am doubtful about letter forms.  There
is also A. S. Kronrod's dictum "Chess is the Drosophila of artificial
intelligence" which may still come to life if more of the chess
programmers regard their activity as a scientific experiment rather
than just as a sport.

The chain of equalities Hofstadter lists and criticizes needs to
be replaced by a chain of relations other than equality if one
wants to make sense of it.