Botvinnik on Computer Chess
Ilan Vardi
waste of time. It goes to show that just because you are good in one
area it doesn't necessarily carry over into another, but that a lot of
people will give you the benefit of the doubt.
Obviously, Botvinnik's plan to write programs which use only narrow
lines has been completely refuted. But the attempt to have
mathematicians analyze geometrical patterns on the chessboard is
ridiculous, especially since he completely ignored the game of GO
where this is much more natural. I believe that the fact that
Botvinnik ignored GO is due to his strength in chess and this
interfered with the purely scientific question of how best to solve
problems in computer science.
Actually Botvinnik's understanding of science is pretty poor. He was
the first to try a ``scientific'' approach to chess, but Emmanuel
Lasker, the best scientist among top level chess players, played it
for the human element.
-Ilan Vardi
Jeff d'Arcy
> Obviously, Botvinnik's plan to write programs which use only narrow
> lines has been completely refuted. But the attempt to have
Bull! I don't know about you, but I'd get pretty antsy if I were in
a position of having to *prove* this claim. I don't think much of MB's
chances here, but we can't just dismiss his attempts on an a priori
basis.
Jeff d'Arcy jda...@encore.com (508) 460-0500
Encore has provided the medium, but the message remains my own
uncia uncia
> Obviously, Botvinnik's plan to write programs which use only narrow
> lines has been completely refuted.
hmm. if i remember correctly, botvinnik's work was done 20
years ago on a machine which ran at an approximate speed of
1.5 Mips, which with his algorithms resulted in a search speed
of about 300 positions/second. this was sufficient to compete
quite favorably against the "brute force" systems of the time,
running on hardware with similar performance, which scored
1000+ positions/second.
given that botvinnik's approach has never been implemented
(it is so much more complicated :-) using "state of
the art" technology (i.e. supercomputers or dedicated chess
machines which enable brute force systems to score ~10K
to 100K+ positions/second), i think such a glib dismissal
of the "heuristic" approach is a bit premature.
--
Michael Collier University of New Mexico Computing Center
col...@ariel.unm.edu 2701 Campus Blvd.
(505) 277 8039 Albuquerque, NM 87131
(Home: 1160 Don Pasqual NW
Los Lunas, NM 87031)
...!cmcl2!beta!unm-la!unmvax!charon!collier
Tom Tedrick
>waste of time.
Oh sure. Joe Random casually dismisses world champions as idiots.
Botvinnik's contributions are profound, and the fact that you don't
have a clue reflects badly on you rather than Botvinnik.
Last time around you were trashing former world CC champion Berliner.
I think one would be better advised to listen carefully to someone
who has proven he can win world-class competitions, rather than
casually dismiss their ideas. They must be doing something right.
**********************************************
High Priest to students: "The facts contradict
our theory. Therefore, the facts must be wrong."
Kurt Luoto
>Botvinnik's whole involvement with computer chess seems like a total
>waste of time.
For that matter, some (such as Albert Einstein) have said that chess in
general is a total waste of time. Are you flaming Botvinnik because
of his hobbies?
> [...]
>
>Obviously, Botvinnik's plan to write programs which use only narrow
>lines has been completely refuted. But the attempt to have
>mathematicians analyze geometrical patterns on the chessboard is
>ridiculous, especially since he completely ignored the game of GO
>where this is much more natural. I believe that the fact that
>Botvinnik ignored GO is due to his strength in chess and this
>interfered with the purely scientific question of how best to solve
>problems in computer science.
Could it be that Botvinnik was simply more interested in chess than in go?
To each his own I say.
>Actually Botvinnik's understanding of science is pretty poor. He was
>the first to try a ``scientific'' approach to chess, but Emmanuel
>Lasker, the best scientist among top level chess players, played it
>for the human element.
Its funny that you refer to Lasker as a scientist. I seem to recall from
my reading that he was a philosopher by trade. I also recall that in an
in interview, he expressed his negative opinions about Einstein's theories
(relativity and all that), and in the process showed an embarrassing lack
of scientific knowledge.
>-Ilan Vardi
Kurt Luoto
Louis Blair
excerpts from a review by Hans Berliner of
Botvinnik's 1983 book, Computers in Chess -
Solving Inexact Search Problems:
Botvinnik is a former World Chess Champion (13 years worth), and a
person who has been highly honored in the Soviet Union for his
contributions to electrical engineering in the power industry there. He
knows about computers and he certainly understands a great deal about
how he himself thinks in producing a chess move. All this appears very
favorable for developing a computer program to play chess in the style
of a human Master, and it is. But, the World has now been waiting for
16 years since Botvinnik's earlier publication (Computers, Chess and
Long-Range Planning) for some result, no matter how intermediate, of
this approach. ... this book does not move the project even one
micro-inch in that direction. ... It talks about the PIONEER program
that plays chess, but does not offer a single game it ever played. ... I
have a lot of sympathy with what Botvinnik has chosen to dedicate the
remainder of his life to: the making of a machine World Chess Champion.
However, he outlined most of that in his earlier book, which gave a
very interesting set of 60 problems that such a machine should be able
to solve. In this volume we are treated to the solution trees to two of
the less difficult of these, together with one other problem that has
somehow wedged ahead of the other 58. ... an examination of the solution
trees that are presented gives the appearance of careful manicuring of
the solution. Not only does the program allow illegal moves to be
tried, it completely ignores moves that a program that uses PIONEER's
avowed search discipline would have to investigate. And most damaging
of all, there is no commentary to tell you why certain branches were
extended and others truncated. This is essential to any understanding
of the method, if indeed there is one. To me this book appears to be a
translation of an updated prospectus prepared 10 years or so ago by
Botvinnik for the Soviet bureaucracy in order to allow him access to a
computer to test his theories. ... When one considers that said
bureaucracy has now taken Botvinnik's computer away for lack of
progress, the value of the document becomes much clearer also. ...
Multi-level control structures such as those Botvinnik advocates have
been tried in this country with notable lack of success, but at least
they worked. The most obvious case that comes to mind is a speech
understanding system that was built here at Carnegie-Mellon University
using such an approach. It required over 20 man-years to build and then
performed worse than another system, also built here, that had a much
simpler control structure and took only 2 man-years to complete.
And here is what Bovinnik himself had to say in an interview that
appeared last year in the ICCA Journal:
Our own program is completely ready on the conceptual level. However,
the coding has been done by mathematicians who do not play chess.
They've produced a very complex program, and there are lots of mistakes
in it. It's about 60,000 lines of FORTRAN. And the computer is a weak
one. ... except when bugs crop up, the program already performs like a
master in the way it constructs a tree of variations. ... It's not in a
position to [play games], because it contains a lot of errors. ...
Getting computing time in Moscow is a big problem. We can use the
machine on Sundays and at night-time, but that's not the proper time for
scientific work. ... At present we get sixteen hours [of computer time]
per week ... A single tree of variations will take up severals hours,
perhaps six or seven. Then you'll clear up one bug, and it's a week
before you can test it again. ... the program was put together in
stages, and contains a lot that is redundant - too complicated. It
could all have been done much more simply. ... We could now rewrite it
from one end to the other, ... but we can hardly undertake this without
a personal computer. ... there's been a crisis in our team - as we
haven't got our computer - two of the four mathematicians who were with
us have left. ... Shubakov and Reznitsky are carrying on. But they'll
go too if we don't get a computer.
Louis Blair
in the ICCA Journal last year:
I believe there were four masters to whom chess-players
are especially indebted. ... I'm talking about the four who
created the "algorithm" for a chess master. First, there was
Greco. Four hundred years ago, he introduced combinations
into chess. Before his time, there had been no such thing. It
was a major step forward. ... The second player in question
was Philidor, 200 years ago. He drew attention to the assess-
ment of positions that are determined by the pawn structure.
The third player was Morphy. He demonstrated positional
understanding in open games, where the pieces are mobile
and the pawns play a subordinate role. What pawns were to
Philodor, pieces were to Morphy. And then Steinitz contributed
an understanding of closed positions. Afterwards, all these
ideas were developed. But in the last hundred years there has
been nothing fundamentally new.