COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	@make(letterhead,Phone"497-4430", Who"John McCarthy",Logo,old,department csd)
C00009 ENDMK
C⊗;
@make(letterhead,Phone"497-4430", Who"John McCarthy",Logo,old,department csd)
@style(indent 8)
@begin(address)
Mr. Lars Kahn
National Swedish Board for Technical Development
Box 43200
100 72 Stockholm
SWEDEN
@end(address)
@greeting(Dear Mr. Kahn:)
@begin(body)        
	I have read the draft report "Contributions to Automatic Programming
Theory".  It was easy to read, because the author writes clearly, even
though he makes many spelling and grammatical mistakes.  I have the
following comments:

	1. The author has read many books and papers, and most of his comments
about them are well taken.  Some I disagree with.

	2. His presentation of the general material is rather superficial,
and one could not learn much about the technical content of the works
from his presentation.

	3. His own contributions (apart from expressions of opinion)
 begin about page 150 and center on
some versions of fuzzy set theory and fuzzy predicate logic.
Most of the results consist in showing that the fuzzy systems
share certain properties of conventional systems.

	4. These versions are rather elementary, and I am not
acquainted enough with previous work in fuzzy logic to determine
their novelty.

	5. Everything was quite clear up to page 180, where I found
some definitions obscure, and this made the rest of that chapter
somewhat unclear.  However, it was still clear that conventional
properties of his new system were being asserted.

	6. The author proposes to use fuzzy set theory and predicate
logic to develop systems of automatic programming.  I found it unclear
what he proposes to do, and I found the example on@* 
pages 215-217
unconvincing, i.e., it didn't seem likely that this approach would
help with data structure selection.

	7. In general I have not found any kind of fuzzy logic
helpful.  The problems of ambiguity and ill-definedness are important
for artificial intelligence and possibly for other branches of
computer science.  However, I don't think fuzzy logic is the
right approach, and the developments based on non-monotonic
reasoning offer greater chances.  I suppose that most people in
AI share my unenthusiasm for fuzzy logic, but it is certainly
an active theoretical field.

	8. I could not imagine it as reasonable to support more
than the author's own time over the next year unless there are
other people eager to work on this project and who have a clearer
idea of what they would do than is expressed in this report.
@newpage

	9. I doubt that the author should be encouraged
to continue in this vein.  Rather he should probably be urged to
specialize his efforts and his writing about the results of his
efforts.  Methodological papers, even by people with great
reputations, rarely have substantial effect.

	10. Unless you find an expert opinion that differs from
my own about the relevance of fuzzy reasoning to automatic
programming, I would suggest you drop the subject.  It seems to
have little connection with the other activities in the UPMAIL
environment - for which I have a much higher regard.

	Let me now answer the three specific questions:

	1. I found the expository sections interesting and
readable but superficial.  The technical section consisted in
showing that certain fuzzy logical systems shared many properties
with conventional logical systems.  Whether this is a contribution
to fuzzy logic, I leave to experts in that field.  However, I
have my doubts about whether fuzzy logic as a whole is a contribution.

	2. The work is not ready for experimental study unless
there are important ideas of which there is no hint in the report.

	3. Yes, I believe that knowledge-based automatic and
construction of computer software is
feasible and that present research in that direction has a good
chance of making important contributions.

	I hope these comments are helpful.
@end(body)
Sincerely,




John McCarthy
Professor of Computer Science