Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Log from Game 2 -Kasparov vs Deep Blue after 35. Bxd6

Author: Rolf Tueschen

Date: 17:23:21 07/27/02

Go up one level in this thread


On July 26, 2002 at 20:42:02, Mike Byrne wrote:

>[D] r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 1
>
>a lot of this is gibberish to me - but I was able to make out that 36.Qb6
>started to fail low and DB went into a "panic mode" and it was the move 36.axb5
>that recovered the score - directly from IBM's site
>
>http://www.research.ibm.com/deepblue/watch/html/game2.log
>
>the last Qb6 pv given was "qf2b6 s.Qe8e7 @s.pa4b5P @s.Ra8b8 @s.qb6a6P @s.Pe5e4
>@y.bc2e4P @s.Qe7e5 y.be4f3 .Rc8d8 .qa6a7 .Qe5c3p .bf3h5".
>
>[D] 1rr3k1/6p1/Q2b1p1p/1P1P1P1B/1Pp5/2q4P/R5P1/R5K1 b - - 0 1
>
>which is really not that great for white ...

This is exactly what Kasparov was thinking. Therefore he tried this, assuming
that DB2 wouldn't find it. DB2 played a different move. Then K asked questions.
IBM refused to show the logs. Then K was out of the match - psychologically.
The the final press conference. K repeated his questions. Answer the logs must
be worked over to be understandable for non-experts.

Then - attention - DB2 was "deconstructed".

So, never more we could research such positions. What is DB2 able to find? We
will never know. Is DB2 able to refuse to go for the material if the
disadvantage becomes only clear at move xy? We will never know.

Dr. Hyatt is right when he insist that we could _never_ know if the given output
is koscher. So deconstruction is without consequences.

Here is the moment for my objection.

It is true that we will never know if the output of parallel computers is
authentic because we can't simply "repeat" the situation. But what "we" could
still do is doing some research with typical chess positions. And Kasparov would
surely be the expert to find interesting positions where a computer, even DB2
would fail to find the correct line. So, in the end we would have a specific
probability for DB2 in certain positions. _Then_ we might get a conviction about
the position of our discussion, where Qb6 was refused and axb5 was played.

For me it's not sound to bash Kasparov for his basic questions and on the other
side quickly minimalize the consequences of the deconstruction. It simply looks
odd. Even if the machine had been sold before. NB here was Kasparov, the best
human chessplayer, asking questions! He should be accepted as a chess expert.

BTW he was invited by the DB2 team exactly for this very reason. Now, something
does not fit here. Ok, Dr. Hyatt once said that after the insinuations by K he
could not expect to get answers. But we know that the insultive aspect was _not_
what Kasparov had expressed, it was already interpretation by the DB2 team, here
in person of M. Campbell who defended against K in a famous press remark after
game 2.

So, we still have the situation, that very basic chess question were asked, who
can't be answered now because the hasty deconstruction. The logs _alone_ are
_never_ a proof as such because they _could_ be doctored in minutes, this is
exactly what Bob told me!

(I want to add that I report, I ask questions, I make conclusions, but I did
never accuse the DB2 team of cheating. That would only be possible if it had
been proven. But this can't happen by definition. So please nobody should read
something into my articles what is not there.)

Rolf Tueschen



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.