Carnegie-Mellon University
Department of Computer Science
Press Release



Date:           27 March 1977
Event Date:     1 April 1977 [Friday] at 12:30 pm EST
Release Date:   Immediate
Subject:        Chess match: Computer vs. International Master
Contact:        Brian Reid, at (412) 621-2600, ext. 141



On Friday, 1 April 1977, at Carnegie-Mellon University, Scottish Chess
Master David Levy will match wits with a computer in a chess match nine
years in the making.

In August of 1968, Levy, a chess player who subsequently attained the
rating of International Master, made a bet with four computer scientists
that no computer would be able to beat him in a chess match within the
next 10 years.  That 10 years will be up soon, and no match has yet been
played.

Chess games between people and computers are not new, but until quite
recently there was no computer chess-playing program that could
consistently beat players of tournament quality.  A chess program called
CHESS 4.5, written by researchers at Northwestern University and running
on Control Data's giant CYBER-176 supercomputer, last month took first
place in the Minnesota Open chess tournament against a field of Class-A
and Expert chess players.

Prof. Donald Michie of the University of Edinburgh in Scotland, one of
the four computer scientists involved in the orignal wager, is visiting
at Carnegie-Mellon this semester.  Buoyed by the success of this
combination of the world's fastest computer and the world's most
successful chess-playing program, Michie invited Levy to come to
Pittsburgh and play a chess match with the machine.  Levy accepted the
challenge, and the match will be played in Carnegie-Mellon's Computer
Science Department on April 1.

Levy will be in a quiet room, alone with a computer terminal, to play
his game.  A closed-circuit television system will bring the results to
spectators elsewhere in the building.  Carnegie-Mellon's Dr. Hans
Berliner, himself an International Master and former world
correspondence chess champion, will provide running commentary and
explanation of the game.  Two of the other computer scientists involved
in the original wager, Prof. John McCarthy of Stanford University and
Prof. Seymour Papert of M.I.T., will be kept up to date via a computer
message system.  Levy's computer terminal will be connected via a
special telephone circuit to the computer at Control Data headquarters
in Minneapolis.

                               BACKGROUND

CHESS 4.5, written by David Slate and Larry Atkin of Northwestern
University, has recently played 16 games in three chess tournaments
against human players; it won two of those tournaments.  This
performance entitles it to a United States Chess Federation (USCF)
rating of 2070, which falls in the "Expert" category.  By comparison,
Levy is rated 2375, and Bobby Fischer is rated 2800.  CHESS 4.5 has
beaten players with ratings as high as 2100, and Levy has lost to
players with ratings lower than 2100, so it is not a foregone conclusion
that Levy will win.  The rules of the bet are that if Levy loses or
ties, he is entitled to a rematch, but if Levy wins, the computer is not
entitled to a rematch this session.  Many thousands of dollars are
involved in the wager, which Levy stands to win if he does not lose to a
computer before August of 1978.  Dr. Berliner believes that the odds of
the computer winning are about one in ten.

Computer scientists study chess and write chess programs as part of the
field known as Artificial Intelligence (AI).  AI researchers study chess
for the same reason that geneticists study fruit flies: it is a
comparatively simple and well-understood problem that can serve as a
test-bed for new techniques in computer problem-solving.  Chess has been
the subject of intensive human study for hundreds of years, and more is
known about it than many other more complicated fields of endeavor.
Carnegie-Mellon is one of the world's major AI research centers, and
many of the techniques used by CHESS 4.5 were pioneered at CMU in the
late 1950's.

The techniques used by CHESS 4.5 are not so much innovative as they are
the result of combining careful refinement of technique with a
tremendously fast computer.  The essential methods used by CHESS 4.5
have been known for years: a computerized representation of knowledge
about chess, and various techniques for using that knowledge to play the
game.  Analogously, the aerodynamic principles used in designing the
Concorde supersonic transport have been known since the late 1940's, but
only recently, by combining careful engineering with a super-powerful
jet engine, were they used to create the first supersonic passenger
plane.

In analysis of the playing record of CHESS 4.5, Dr. Berliner notes that
its tactics are superb, of Grandmaster quality, but that its strategy
and long-term planning ability are only mediocre.  As faster computers
are built and better strategies are developed, the program can only get
better.