COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input memo.tex[let,jmc]
C00005 00003	\smallskip\centerline{Copyright \copyright\ \number\year\ by John McCarthy}
C00006 00004	%logic[w89,jmc]		Logic in AI, for csli
C00008 ENDMK
C⊗;
\input memo.tex[let,jmc]
\title{Logic in Artifical Intelligence---Main Points---1988 Jan 25}

	In this talk I want to explain the approach to AI via
math logic.  The main content will be examples of formalizing
phenomena that might be considered problematical - illustrating
my hopes.

0. We regard AI as studying successful behavior in complex
informatic situations.  We are not especially trying to imitate
human physiology or psychology.  AI is a branch of computer
science---not a branch of biology.

1. Human-level AI systems face the {\it common sense informatic situation}.
Decisions must be made where there isn't a mathematical model
of the relevant information.  It isn't known in advance what
information is relevant.  This precludes a conventional scientific
theory.

2. Main idea: Use logic to express what an AI system knows about
the world and use logical inference to decide what actions will
achieve its goals.  The program that does the inference should be
simple and standard.  The intelligence is embodied in the facts.

3. Softening of main idea: Use programs where available, but reason
about the results of the programs.  The logic is boss.

4. Inference is not just deduction.  It turns out that nonmonotonic
reasoning is useful (and human).  Mathematical logic provides
the tools for formalizing nonmonotonic reasoning.  (Contrary to
some previous intuitions).

5. Examples of formalization: blocks world, knowledge up to Mr. S
and Mr. P puzzle, inheritance with exceptions.

6. Tools: situation calculus, circumscription, formalization of
context.

7. Relation to other scientific activities: philosophy, philosophical
logic, natural language.

8. Push Dennett's design stance.
\smallskip\centerline{Copyright \copyright\ \number\year\ by John McCarthy}
\smallskip\noindent{This draft of logic[w89,jmc] TEXed on \jmcdate\ at \theTime}
\vfill\eject\end
%logic[w89,jmc]		Logic in AI, for csli

	In this talk I want to explain the approach to AI via
math logic.  The main content will be examples of formalizing
phenomena that might be considered problematical - illustrating
my hopes.

Preliminaries
AI as cs
1959 paper
logic programming
epistemological adequacy in the common sense informatic situation
Russell's mistake
various clarifications

major topics
nonmonotonic reasoning, birdcage example
ambiguity tolerance
approximate theories
context
meta-epistemology? - make this last?

contacts with other subjects
philosophy
philosophical logic
natural language - Insufficient use has been made of the fact that
natural language works in the common sense informatic situation.

all via design stance
	1. what you have to know about it to use it
	2. how you can use the fact that it's got to work
to learn more about it.  When should a robot use ``but'' in communicating
with another robot?