COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\|\\M1BASL30\M2NGR40L\M3NGR25\M4NGR20\F2\CSTANFORD UNIVERSITY
C00009 ENDMK
C⊗;
\|\\M1BASL30;\M2NGR40L;\M3NGR25;\M4NGR20;\F2\CSTANFORD UNIVERSITY
\F3\CSTANFORD, CALIFORNIA 94305

\F4ARTIFICIAL INTELLIGENCE LABORATORY\←L\-R\/'7;\+R\→.\→S   Telephone:
\←S\→.415-497-3330

\F1\CMARCH 31 1977


Mr. Eric Jensen
Graduate School of Business
Duke University
Durham, North Carolina  27706


Dear Mr. Jensen:

\JI was very sorry to have had to disappoint you the other day.  The trouble
has been found and we will be able to have our match just as soon as I get
back in town.

I have two versions of my program and I was trying to use the learning
data for one program with the other one.  As a matter of fact, the
learning data for my most recent version seems to have been lost and I now
have two copies of the older data, one copy mislabeled.  The older version
was nearly as good as the newer one, its principal failing being in the
absence of some operating refinements that have been added to the newer one.

I do have some suggestions as to protocol which you might consider.

In the first place, if we play only one game each of us will be sitting
around, or rather our computers will be so engaged, while we wait for the
other machine to play.  It seems logical, therefore, for us to play two
games simultaneously.

So then the following suggestion:  We each start a game with our own
program playing black.  As a means of identification the game started at
Stanford will be called the Stanford game and the one started at Duke will
be called the Duke game.  As soon as both machines have moved, we exchange
moves and start our programs on the replies for the interchanged games.
This process can then be repeated.

Will this be difficult for you to do?

Here at Stanford, I can simply map
my console to another one and then switch back and forth between games
without in any way disturbing the state of the other game.  I could
arrange to have two consoles in the same room but this would require some
special wiring.

I am of two minds as to whether or not I should allow my program to use book
information that it has stored away.  On the whole, I am inclined to use this
information.  People certainly use remembered information when they play, and
the accumulation of this information and its storage and cataloging  so as to
make it quickly available was not a easy thing to do, at least it was not easy
at the time when I did it.  Now-a-days there are all sorts of file retrieval
systems which tackle this problem.  Without this stored data, my program is
painfully slow at the start of a game and it still almost always plays a book
move but only after a long delay.

As to whether we should have a clock on the game,-  I can do almost anything that
you think desirable.  At the moment, there is something wrong with the run-time
clock in our computer but I am assuming that this will be fixed in the very near
future.  I am, however, unable to check out the time control features in the
program and there could be something wrong here.  If we really want a comparison
between the programs independent of the machines on which they are run, I should
be allowed longer in proportion to the speeds of our two machines.

I judge that you simply want a rough measure and all of this refinement is
hardly necessary.  I will be very surprised if your program does not play
rings around mine in any case.  After all it has been over 6 years since I
have even had mine operating.

The incentive of this projected match has forced me to get my program back
into an operating state, under a different operating system, using a newer
editing program with a different compiler and for scope displays rather than TTY's
for which it was originally written and lastly with the older debugging aids
written into the program no longer usable with our present system.  I should
have done all of this a long time ago.\.

\←L\→S\←R\-L\/'2;\+L\→L



Sincerely,




Arthur L. Samuel



\←S\→L
ALS:pdp10