perm filename AICIRC.ABS[CUR,JMC] blob sn#144963 filedate 1975-02-07 generic text, type C, neo UTF8
COMMENT āŠ—   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M0BASL30\M1BASI30\M2NGR40\F2
C00004 ENDMK
CāŠ—;
\\M0BASL30;\M1BASI30;\M2NGR40;\F2
\CEPISTEMOLOGY, FORMAL REASONING AND AI
\F0
\J	AI, like any difficult science, benefits from identifying subproblems
that can be attacked separately and whose relation to the whole problem is
sufficiently well defined that progress in the subproblem will be genuinely
relevant to the whole problem.  Such is the \F1epistemological\F0 part of the AI
problem, namely expressing facts about the world in formal language
and expressing the reasoning whereby facts about how to solve a problem
follow from its description.
The \F1heuristic\F0 part of the problem is concerned with search and
every heuristic program has an epistemological base, usually naively
chosen.

  For epistemology, first order logic with
set theory and fortified with reflection principles and the ability to
refer to the results of running programs on complex data structures is
suitable.  However, there are many ways of using first order logic, and
almost all of them are wrong.  Examples will be given.\.