perm filename GURBA.1[LET,JMC] blob sn#720296 filedate 1983-07-22 generic text, type C, neo UTF8
COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub[let,jmc]" source
C00006 ENDMK
C⊗;
.require "let.pub[let,jmc]" source;
∂CSL Mr. Krzsstof Gurba↓Dept. of Philosophy↓Jagiellonian University↓
Wisniowa 20/115↓Krakow↓Poland∞

Dear Mr. Gurba:

	I'm sorry that I don't have time to think about your letter
in time to answer it more thoroughly.  As it happens, naturalness
as a predicate does not correspond to the way I think about
knowledge representation and natural language.  Nevertheless, your
letter does raise some questions to which I can react.

	In my opinion, natural language has many features which
are necessary for effective thought and which have not yet been
realized in any languages or representation systems used in artificial
intelligence.  It is not easy to tell what these are.  The enclosed
papers seem to me to be somewhat relevant, and I hope they are
helpful.

	Here are some responses to some of your questions.

2 and 3. The open ended character of human concepts and the way
they permit indefinite elaboration is something that will be required
before AI systems can have general intelligence.

4. The only way I know to identify these features is to design, at
least conceptually, AI systems that require them.  For example,
I am trying to formalize the facts about travel in a way that
admits the addition of new kinds of obstacles and new ways to overcome
them without changing previous common sense axioms - only adding
new ones.  Of course, this requires using non-monotonic reasoning
methods.

5. I don't think physiology is advanced enough to contribute to
the study of intelligence yet, and I think questions of intelligence
can be separated from cultural issues.

6. I follow a top down approach, but I start below the cultural
level at the problem solving level.  However, other approaches
may also be useful.

7. I think we can understand intelligence well enough to produce it
in machines, but it may take a long time.

8. While language can be a weapon, it is powerful only when reinforced
by police.

	I hope these remarks are helpful.

.sgn