Ideas not on slides

	finitization
	dogs and garbage cans

Some expert systems and common sense

	a.  when to use Mycin
	b.  prognosis within Mycin

Apology

	a.  not a full report
	b.  I haven't solved the problem

Logic level

"Perception is inverse computer graphics"
"fractual fus   World built from perception

__________

dogs and cans

1. false solution -- each a finitization
2. prevention
3. repeated events

	First, I want to minimize misunderstanding.  This lecture  is 
primarily addressed to people doing basic research in AI.  It concerns problems
that still offer great difficulties.  It is not a report to the press or public 
on the general state of AI nor is it aimed at helping apply present AI 
technology.

	Common sense has again become a hot topic in AI, at least in Northern
California.  SRI and CSLI are having its "common sense summer" on which
Jerry Hobbs has already reported.

	There are several intuitions behind this activity.  

1.  There is a collection of facts and methods used by humans that would make
possible much more powerful, general and flexible computer programs if AI
included it.

2.  Specialized knowledge and ability need to be incorported in a general 
    common sense framework that [Mycin example] decides when the specialized
    information is appled.  

3.  Many of us also believe that we need something other than 
    adding large amounts of common sense knowledge to the databases of present
    expert systems.        

4.  It is hoped that a more sophisticated knowledge using technology can be 
    developed that would make incorporating a human's common  sense knowledge just
    a matter of adding to a common sense database.  There are no present 
    candidates for such a formalism.

	Some of what I say may be controversial.  However, I will not try to
objections that might be raised to various aspects of this analysis.  Perhaps 
I will do some of it in the published paper.