A special thanks to Harvey Mandell and Aviv Friedman who have contributed to this lecture.
Over the last decade, computer programmers with grandmaster guidance have brought computers to the stage where they beat grandmasters. One of the best programs is the Israeli developed ”Deep Junior” three time and current world computer champion.
Junior’s programmers Shay Bushinsky and Amir Ban are both brilliant programmers but mediocre chess players. I am the grandmaster part of the Junior trio. My role is to aid in the opening preparation as well as to detect the problem of each version Junior programmers are creating and to decide which one is the best.
It seemed that the ultimate test for Junior was to see how it could do against Garry Kasparov, considered by many to be the best chess player of all time. This match took place seven years after the famous match with IBM’s Deep Blue (1997), when the world’s leading player had returned to play in New York once again representing humans against technology. The match was sponsored by FIDE, the international chess federation and X3D technologies.
This article tells some of the inside stories of the match as I saw it by my participation.
Before the beginning of the match we did not expect … any specific result. We – simply hoped that Deep Junior would act adequately. It seemed that everyone except for us was sure of Kasparov’s victory. It wasn’t clear to me why so many had reached that conclusion. In 2002 Kramnik had failed to beat Deep Fritz, using anti-computer strategy. Also, Kasparov’s style was not considered to be the best for a game against the computer.
The venue of the match, an impressive 24 story building which is an all-in-one combination of hotel, health spa, gymnasium, conference rooms, restaurant and a lot more, had dedicated its 12th floor as the playing ground. Blocked off to the general public and manned by no-nonsense security guards, the floor contained a special room for Garry, one for the Deep Junior team where the physical machines rested (and where representatives of the ICGA made sure it was ONLY the computer making the moves) and of course the actual playing room.
The first game effectively changed the situation.
Kasparov,G – Deep,Junior , New York 2003
1.d4 d5 2.c4 c6 3.Nc3 Nf6 4.e3 e6 5.Nf3 Nbd7 6.Qc2 Bd6 7.g4!
The first sign of combat! This move, attributed to GM’s Shabalov and Shirov came as a surprise to me. How to expect such a sharp line when you play against the computer?!
7…dxc4 8.Bxc4 b6 9.e4 e5 10.g5 Nh5 11.Be3 0–0 12.0–0–0 Qc7 13.d5
Definitely that was not the best position Junior got after the Opening.
This just makes things worse… 12…Bb7!? or even 12…c5 were better, although white already has a nice edge.
14.dxc6 bxc4 15.Nb5 Qxc6 16.Nxd6 Bb7 17.Qc3
No other computers would have played this move, but as a computer when the evaluation of losing e5 with a bad position is more than sacrificing an exchange for some ‘life’, you play it.
18.Nxe8 Rxe8 19.Rhe1 Qb5 20.Nd2 Rc8 21.Kb1 Nf8 22.Ka1 Ng6 23.Rc1 Ba6 24.b3 cxb3 25.Qxb3 Ra8 26.Qxb5 Bxb5 27.Rc7
Completely unexpectedly for all of us, Kasparov decided to use all his opening potential, and “to break” a computer quickly in the opening.
The opening stage is in general a rather hard test for the machine and perhaps, its weakest place. In its opening the computer is still not ready to enter serious rivalry with humans.You see on a board – 32 figures, and it is enormously difficult for the computer to expect variants. And very frequently the opening phase puts a computer in a tough situation.
That is why the Opening Book is such an important part of the preparation. The Computer like a human should feel quite confident. Sometimes the situation could be pretty funny. I used to detect many positions, where according to theory the line appears to be a good one, but the Computer feels like it is going to lose.I could compare it to the kid who should taste the ice cream before eating it.
The same thing with the Computer – you have to feed it and to see the feedback, before deciding if the Computer will feel comfortable enough.
So, game 1 Deep Junior lost because Garry found the hole in our Opening Book, but as well understood that the best strategy vs. Junior is to push pawns forward, win a territory and get more space … Junior panicked in such situation, made a few positional errors and Kasparov simply dominated.
Filed under: Kasparov against Deep Junior |