COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub[let,jmc]" source
C00008 ENDMK
C⊗;
.require "let.pub[let,jmc]" source
∂AIL Prof. Andrei P. Ershov↓Computation Center
↓Novosibirsk 630090↓USSR∞

Dear Andrei:

	I'm sorry I couldn't make it to the IFIP conference
this year, and you couldn't come to the AI
conference.  In my opinion, this AI conference
was the best yet, especially in the area of problem
solving systems.

	I have been further pursuing the ideas in the
papers I gave you in Repino, with only small advances
beyond the results reported in the papers.

	I enclose a prospectus for a one year study
on artificial intelligence and philosophy which the
Center for Advanced Study in the Behavioral Sciences
has asked me to organize.  Candidates for participation
include, besides myself, Daniel Dennet of Tufts
University, Margaret Boden of Sussex University
who has just written an excellent popular
but constructively critical book on AI but probably can't
come, and Zenon Pylyshyn of the University of
Western Ontario, an AI oriented theoretical psychologist.
A really outstanding Soviet candidate could
be considered.

	Here is some additional news:

	1. The Fletcher Jones Foundation has endowed a chair at
Stanford, and Don Knuth has been appointed to it.
Jones was one of the founders of an early successful software
company and was killed in an airplane crash.

2. Knuth is spending a good part of his time this year
developing an automatic type-setting system good enough
for his books.  He wants to get a Videocomp typesetter
from III that will permit us to produce good mathematical
quality books.  Winston's recent book on AI was typeset
by the Videocomp.

3. Dialnet is proceeding well.  Here 
is a draft of the low level protocols.  They are subject
to much further discussion before even an experimental
implementation.

3. Dave Poole and friends have finished building
Foonly at III and are now debugging it.

5. Another grooup of Stanford students
have designed a big computer called the S-1.  It is intended
to be a multi-processor machine with each processor about
the speed of a 370/168.  They have or are about to
get enough money for a 16 processor system.
The first processor is about half built.  We are making a proposal
to design and implement a time-sharing system for it,
and would get a system in addition
to money for people.  The machine has
a 30 bit address to the byte level, but it is more
like a PDP-10 in structure than like a 370.  I
think it embodies the right approach to multi-processor
architecture - each processor has independent control and
can address the whole of memory.

6. The LOTS time-sharing system is now working well,
and Ralph Gorin is doing an excellent job as manager.
We now have 512K words of core, 2 disk channels,
and 3 disks each holding 200 megabytes.  It runs very
well with about 45 logged in users, and we will shortly
try it out with about 20-25 more terminals.  We
expect its performance will still be adequate.

7. Rumor has it that the next generation of Intel
and Zilog microprocessors will have large address spaces -
some say up to 32 bits which will provide room for
growth and for upwards compatible machines.  Other
rumors contradict these.  We'll know early next year.

	How are things in Novosibirsk?  Is there any
scientific computer much better than the BESM-6 in
the offing?  How is it with big disks and with time-sharing?

	How are you and how are Nina and Vasya and
Anyuta?  I think you are very fortunate in your
children.

.reg