⬆️ ⬇️

Kasparov vs. Deep Blue. Part II: The Philadelphia Experiment



After a break, I continue a series of articles (and also promise to finish it soon) about the many years of chess struggle between two worlds - the human and the computer. It was planned that Rom77 would write articles about both of Kasparov’s battles with Deep Blue, but unfortunately, my co-author, who wrote a great start , does not get in touch. I sometimes receive letters from grateful readers asking why there is still no continuation, therefore, once again I take up the pen and sit down at the keyboard in order to continue the interrupted story about the adventures of Harry and his silicon friends. In addition to the scattered information about the events of those years, the article also decided to use Roman’s drafts for the 1996 match, which I had at my disposal.





Some dry facts. The match took place in Philadelphia from February 10 to February 17, 1996 (hmm, only 19 years have passed). The prize fund was 500 thousand dollars of which 80% (400 thousand) were received by the winner and 20% (100 thousand) - by the loser.



Basically, in front of Kasparov, Fen Xu sat at the computer, but sometimes he was replaced by Murray Campbell or Joe Hohen . All three are creators of Deep Blue. The decision to stop or continue the game (the delivery of the game, an agreement for a draw) was made by a specially invited chess consultant, grandmaster Joel Benjamin . Read more about the IBM team in the first part .



1st party. Deep Blue - Kasparov. 1: 0





So, the match has begun. In response to the immortal 1.e2-e4, Garik chooses his formidable weapon for the black ones - the Sicilian defense. This acute start has repeatedly brought great victories to the grandmaster, but the fact is that even Kasparov does not know and cannot know the subtle nuances of all the branches of the opening. The computer acted wisely (however, the creators programmed it, apparently, this “wisdom”) —in its debut base, quickly came down from the options preferred by Kasparov, and turned the game into a channel in which Harry never won victories. Black does not manage to equalize the game. And on move 23, the computer broke up the enemy's position by conducting a deafening d4-d5 pawn breakthrough (see diagram).

')



Upon further review of the game, it may seem that the grandmaster managed to develop a counter-matte attack, and who will win - this is the grandmother who said it in two. But in fact, whites from the beginning to the end controlled the situation and were always at the checker step ahead. In the worst position, Kasparov makes a mistake and the computer ends the game with an exponential execution of the black king (in the diagram, the final position of the meeting). Pale Kasparov is quickly removed from the hall ... People cheer - the robot convincingly outplayed the world champion already in the first sparring!



2 party. Kasparov - Deep Blue. 1: 0





We must pay tribute to Garik - he always knows how to mobilize after offensive defeats. Kasparov for whites chooses a solid and reliable option. Black balances have a minus pawn, but they managed to achieve "multi-colored" bishops (see diagram). With a calm position, this allows you to reasonably hope for a draw. But the troubled times on the board - the white king reliably hid behind his pawns, but the dark lord is all in full view and the “colorful” with queens just allows for a very dangerous attack on him. The computer was somehow protected from dull threats, but for this it was necessary to give almost all the remaining infantry. Then there was a queen exchange in the game and, despite the colored elephants, there is no longer any chance of a draw - the string of white soldiers cannot be stopped, but the only surviving black canine was easily blocked.



3 party. Deep Blue - Kasparov. ½: ½





Neither the whites nor the blacks in the game took the risk and the duel went without incident. Kasparov gained some advantage due to the active rook and better knight compared to the enemy bishop. The computer found an ingenious rookie maneuver that finally equalized the odds. Bloodless and quick draw.



So half the match is over. The rivals lost once, but the peaceful third party showed that they quickly recovered after the strikes. The opposing sides decided that it was time again for decisive action.



4 party. Kasparov - Deep Blue. ½: ½





Kasparov played White very actively, but Black built an impregnable fortress. The world champion offered a draw, but the IBM staff decided that they could hope for more than a peace agreement. Then Kasparov gave up “quality” and created a position (see diagram), where black, despite the “extra” rook, has no chance of winning - it’s dangerous to move far away from his king, it’s time to mate.



Not agreeing to a draw right away, in this game the Aybians didn’t particularly risk anything. In the next game, in another drawn position, they decided to repeat this “maneuver” again. But as it turned out - with more fatal consequences.



5 party. Deep Blue - Kasparov. 0: 1





The computer again played reliably and Kasparov, despite the fact that he was able to seize the initiative in the party in black, did not achieve something special. The world champion offered a draw (see diagram) and again the IBM team decided that it was possible to play some more. Then Kasparov increased the activity, but the car played apathetic and unprincipled. Harry led the e- pawn forward and White had to give the piece for it. The man took the lead.



It was a turning point of the match. When Harry offered to disperse the world, opposite him at the computer was, as usual, Fen Xu. The "father" of a chess supercomputer in the game was understood much less than in the high-speed processors of his brainchild. Therefore, the decision was given to Grandmaster Benjamin. Joel realized that a draw in this case was a good offer, but decided to continue the game "in the interests of science." Later he explained at a press conference that he wanted to observe how the car would behave in a roughly equal position. Kasparov heartily thanked for the experiment.



6 party. Kasparov - Deep Blue. 1: 0





Inspired by the success, Kasparov on the rise spent in white the last batch, having conducted a series of subtle, truly "human" maneuvers. The computer was doomed to play "from defense." But while the car was building fortifications on the last frontiers, the man seized the space and maximally expanded the sphere of influence of his figures. In the diagram - the final position of the party. On the board is the so-called tsutswang in the middle game. With a full board of pieces, black has almost no moves. Virtually any movement follows, if not forced mate, then major material losses. The computer now has even an extra pawn, but White's position is so overwhelming that the Deep Blue team decided: there is no point in delaying the end.



At the post-match press conference, pleased Harry Kimovich did not stint on the praise of the electronic opponent and its creators:



"First of all, I would like to congratulate the IBM team on truly outstanding achievements, thanks to which the computer turned out to be such a serious opponent."


"This happened for the first time in history - so that a computer with a chess program created by people, most of whom do not know how to play chess, fought so seriously with the world champion."


Also interestingly commented on the computer damage in the last two installments:

“Perhaps those who were disappointed with the 6th or 5th games can say that in many positions the machine played rather ineptly, being unable to understand the simplest things that are easily assimilated by a human chess player. But, firstly, such positions needed to be able to create, and secondly, it was necessary to constantly monitor tactical strikes that can instantly decide the outcome of the game — this is one of the important advantages of the machine. ”


After that, Kasparov shared how he managed to find the key to the electronic opponent:

“I think that the main difference between a person and a computer can be felt by studying the 1st and 2nd installments. I played the last two games, learning to play the positions and choose the moves that are most unpleasant for the car. For computers, the learning process is quite difficult, and during the match it is generally not possible, since the team needs time to conduct the learning process under laboratory conditions. ”


“It was fundamentally important that I very quickly managed to learn and chose the right strategy. In fact, I avoided what was to be avoided. Already only one first batch gave me enough information. Do not create weaknesses, do not play open positions where there is a chance that your king may be in danger. And I in the next five games avoided a direct clash. The positions were open, but still the ability of the machine to create something unpleasant for my king or other figures was very, very limited. ”


And finally, he predicted a great future for computer chess:

I believe that for the first time in the history of mankind we have something very close to artificial intelligence. It is already possible to reduce the number of errors and reach the level of the game, which is determined by the intuition of the person. Now, where I intuitively feel, whether I play correctly or erroneously, the machine simply goes through billions and billions of options. But it is possible that in the future it will be possible to create an algorithm that would be comparable with the intuitive thinking of a person.




Kasparov's victory was quite convincing in the long run (two points advantage), but there was no overwhelming advantage in the game. In fact, man dominated the machine only in two batches of six - the second and last. The computer had an unconditional advantage in the first game, so a more “fair" score in terms of the number of victories would be 2: 1, not 3: 1.



IBM decided to re-throw the glove to the champion, offering to play the same match a year later. Everyone understood that this year would be spent on improving the algorithms, taking into account the experience of the match, increasing the counting power, as well as further “honing” the game of the supercomputer under the style of Kasparov. Most experts agreed that the second 6-round fight will also remain for Garik, although the victory will be tough. And this is despite the fact that on the eve of the first match the supercomputer was considered the favorite.



But, to everyone's surprise, Deep Blue will unexpectedly take revenge in very controversial circumstances ...



( To be continued ... )



Additionally



Match table

one23fourfive6ResultGlasses
Garry Kasparov0one½½oneone+3, –1, = 2four
Deep blueone0½½00+1, –3, = 22


Debut statistics



Sicilian Defense - 1, 3

Catalan beginning - 2

Slavic protection - 4, 6

The debut of four horses - 5



View the party in detail here .

And about Deep Blue there is a detailed Wikipedia article .



All articles of the series
Kasparov - Deep Thought. One gate play

The first offenders. Fritz and Genius

Kasparov vs. Deep Blue. Part I: Black Box

Kasparov vs. Deep Blue. Part II: The Philadelphia Experiment

Kasparov vs. Deep Blue. Part III: Mezhmatche

Kasparov vs. Deep Blue. Part IV: New York Secrets

Kasparov vs. Deep Junior. Return to New York

Chess 3D-show: Kasparov VS Fritz

After Kasparov. Computer chess - results and prospects

Source: https://habr.com/ru/post/365045/



All Articles