Retour au menu

Matchs entre ordinateurs et champions du monde

[lang_fr]

Kasparov – Deep Blue

Deep Blue est un superordinateur spécialisé dans le jeu d’ échecs par adjonction de circuits spécifiques, développé par IBM au début des années 1990 et qui a battu le champion du monde d’échecs Garry Kasparov en 1997, mais en dehors des conditions exigées lors des championnats.

 

Histoire

DEEP BLUE :

C’est l’aboutissement du projet ChipTest, qui fut lancé par les étudiants Feng-hsiung Hsu, Murray Campbell et Thomas Anantharaman au laboratoire de l’ Université Carnegie Mellon en 1985. Renommé Deep Thought en 1988, le projet est devenu Deep Blue en 1993.

Deep Blue a rencontré le champion du monde d’échecs du moment, Garry Kasparov, le 10 février 1996. La machine remporte la première partie, mais Kasparov remporte trois autres parties et le match se solde par 4-2 en faveur du champion. Deep Blue de 1996 mesurait deux mètres de haut et pesait 700 kg. Il s’agissait d’un super-calculateur IBM ( RS/6000 Scalable POWER parallel Systems) dont chacun des 32 processeurs consacrés au calcul pur a été connecté à une carte comprenant 8 processeurs dédiés aux échecs, soit au total 256 processeurs spécialisés fonctionnant en parallèle.

 

DEEPER BLUE :

En mai 1997 a lieu le match revanche en six parties entre Deep Blue (parfois surnommé Deeper Blue à cette occasion) et Kasparov. Pour la première fois de l’histoire, le champion du monde dut s’incliner contre un ordinateur, sur le score de 2½ à 3½. En 1997, Deep Blue pesait 1,4 tonne et mesurait 1,80 m. Il fallait 20 personnes pour qu’il fonctionne. En juin 1997 Deep Blue occupait la 259^e place au TOP500 des supercalculateurs et avait une puissance de 11,38 GFLOPS.

Deep Blue calculait alors entre 100 et 300 millions de coups par seconde, il pouvait calculer de 12 demi-coups de profondeur en moyenne. Les grands maîtres d’échecs Miguel Illescas, John Fedorowicz, Nick De Firmian et Joel Benjamin aidèrent à sa conception, notamment de sa bibliothèque d’ouverture.

Cette victoire a été contestée par Kasparov, car l’équipe de programmeurs de Deep Blue avait accès à toutes les parties de Kasparov, tandis que celui-ci n’avait pas eu accès à la liste chronologique des parties jouées par Deep Blue. Kasparov a également affirmé que certains coups étaient l’œuvre d’un grand maître, et demanda, mais n’obtint pas, les algorithmes de calculs de variantes du logiciel.

Une autre cause est à chercher dans la stratégie de Kasparov, qui s’est écarté rapidement des sentiers battus de la théorie dans certaines parties, pour éviter les stratégies préétablies à l’avance, ce qui s’est malheureusement retourné contre lui dans la dernière partie.

Grâce à son incroyable force de calcul, Deep Blue a joué lors de ce match quelques coups de la classe d’un grand maître, dont l’un, abondamment commenté dans les revues spécialisées, a complètement déstabilisé Kasparov. Il n’a pas cependant complètement convaincu de la supériorité des logiciels d’échecs, comme le montre la première partie.

En dépit des demandes de Kasparov, il n’a pas été possible d’organiser un match revanche, l’ordinateur ayant été démantelé et ses concepteurs assignés à d’autres projets.

Match de 1996

Première partie

Deep Blue – Kasparov 1996, 1^re partie ¹

Deuxième partie

Kasparov – Deep Blue, 1996, 2^e partie ²

Match revanche de 1997

Partie 1 : Deep Blue n’a pas commis d’erreur flagrante, mais a joué des coups dépourvus de vision stratégique, ce qui a permis à Kasparov de gagner la partie facilement. Cette partie illustre bien l’effet d’horizon et l’absence de plan stratégique typiques de la façon de jouer des logiciels d’échecs.

Partie 2 : À l’inverse de la première, la position est plus ouverte et c’est la force tactique de la machine qui a parlé. Kasparov a été très surpris par le 36^e coup 36.axb5!? car le programme évite de capturer un pion et empêche ainsi Kasparov d’obtenir une attaque sur le roi ennemi. Kasparov abandonne prématurément après 45 coups, car après analyse il rata un coup menant à un échec perpétuel que la machine ne vit pas.

Partie 3 : Kasparov assure une nulle.

Partie 4 : Kasparov ne parvient pas à avoir un avantage tangible, et la partie aboutit à une finale de Tours nulle après 5 heures de jeu.

Partie 5 : Une finale Tour + Cavalier, avec un pion passé pour Kasparov, mais la partie était nulle par échec perpétuel.

Partie 6 : Kasparov, fatigué physiquement et psychologiquement, sort de son répertoire d’ouvertures, ce qui permet à Deep Blue de placer un sacrifice gagnant dans l’ouverture, et forcer le champion du monde à abandonner après 19 coups à peine. Certains affirment qu’il ne perdit pas cette partie parce que la machine était plus forte que lui, mais parce qu’il était trop fatigué.

Meilleures parties :

Première partie

Kasparov – Deep Blue, New York 1997, 1^re partie

Deuxième partie

Deep Blue – Kasparov, New York 1997, 2^e partie

Sixième partie

Deep Blue – Kasparov, New York 1997, 6^e partie

[/lang_fr]
[lang_en]

Deep Blue (chess computer)

 

Deep Blue was a chess-playing computer developed by IBM. On May 11, 1997, the machine won a six-game match by two wins to one with three draws against world champion Garry Kasparov. ^[1] Kasparov accused IBM of cheating and demanded a rematch, but IBM refused and dismantled Deep Blue. ^[2] Kasparov had beaten a previous version of Deep Blue in 1996.

 

Origins

The project was started as ChipTest at Carnegie Mellon University by Feng-hsiung Hsu, followed by its successor, Deep Thought. After their graduation from Carnegie Mellon, Hsu, Thomas Anantharaman, and Murray Campbell from the Deep Thought team were hired by IBM Research to continue their quest to build a chess machine that could defeat the world champion. Hsu and Campbell joined IBM in autumn 1989, with Anantharaman following later. Anantharaman subsequently left IBM for Wall Street and Arthur Joseph Hoane joined the team to perform programming tasks. Jerry Brody, a long-time employee of IBM Research, was recruited for the team in 1990. The team was managed first by Randy Moulic, followed by Chung-Jen (C J) Tan.

After Deep Thought’s 1989 match against Kasparov, IBM held a contest to rename the chess machine and it became « Deep Blue », a play on IBM’s nickname, Big Blue. After a scaled down version of Deep Blue, Deep Blue Jr., played Grandmaster Joel Benjamin, Hsu and Campbell decided that Benjamin was the expert they were looking for to develop Deep Blue’s opening book, and Benjamin was signed by IBM Research to assist with the preparations for Deep Blue’s matches against Garry Kasparov.

In 1995 « Deep Blue prototype » (actually Deep Thought II, renamed for PR reasons) played in the 8th World Computer Chess Championship. Deep Blue prototype played the computer program Wchess to a draw while Wchess was running on a personal computer. In round 5 Deep Blue prototype had the white pieces and lost to the computer program Fritz in 39 moves while Fritz was running on a personal computer. In the end of the championship Deep Blue prototype was tied for second place with the computer program Junior while Junior was running on a personal computer.

Deep Blue versus Kasparov

On February 10, 1996, Deep Blue became the first machine to win a chess game against a reigning world champion (Garry Kasparov) under regular time controls. However, Kasparov won three and drew two of the following five games, beating Deep Blue by a score of 4–2 (wins count 1 point, draws count ½ point). The match concluded on February 17, 1996.

Deep Blue was then heavily upgraded (unofficially nicknamed « Deeper Blue ») and played Kasparov again in May 1997, winning the six-game rematch 3½–2½, ending on May 11. Deep Blue won the deciding game six after Kasparov made a mistake in the opening, becoming the first computer system to defeat a reigning world champion in a match under standard chess tournament time controls.

The system derived its playing strength mainly out of brute force computing power. It was a massively parallel, RS/6000 SP Thin P2SC-based system with 30 nodes, with each node containing a 120 MHz P2SC microprocessor for a total of 30, enhanced with 480 special purpose VLSI chess chips. Its chess playing program was written in C and ran under the AIX operating system. It was capable of evaluating 200 million positions per second, twice as fast as the 1996 version. In June 1997, Deep Blue was the 259th most powerful supercomputer according to the TOP500 list, achieving 11.38 GFLOPS on the High-Performance LINPACK benchmark.

The Deep Blue chess computer which defeated Kasparov in 1997 would typically search to a depth of between six and eight moves to a maximum of twenty or even more moves in some situations. Levy and Newborn estimate that one additional ply (half-move) increases the playing strength 50 to 70 Elo points ( Levy & Newborn 1991:192).

Deep Blue’s evaluation function was initially written in a generalized form, with many to-be-determined parameters (e.g. how important is a safe king position compared to a space advantage in the center, etc.). The optimal values for these parameters were then determined by the system itself, by analyzing thousands of master games. The evaluation function had been split into 8,000 parts, many of them designed for special positions. In the opening book there were over 4,000 positions and 700,000 grandmaster games. The endgame database contained many six piece endgames and five or fewer piece positions. Before the second match, the chess knowledge of the program was fine tuned by grandmaster Joel Benjamin. The opening library was provided by grandmasters Miguel Illescas, John Fedorowicz, and Nick de Firmian. When Kasparov requested that he be allowed to study other games that Deep Blue had played so as to better understand his opponent, IBM refused. However, Kasparov did study many popular PC computer games to become familiar with computer game play in general.

Aftermath

After the loss, Kasparov said that he sometimes saw deep intelligence and creativity in the machine’s moves, suggesting that during the second game, human chess players had intervened on behalf of the machine, which would be a violation of the rules. IBM denied that it cheated, saying the only human intervention occurred between games. The rules provided for the developers to modify the program between games, an opportunity they said they used to shore up weaknesses in the computer’s play that were revealed during the course of the match. This allowed the computer to avoid a trap in the final game that it had fallen for twice before. Kasparov requested printouts of the machine’s log files but IBM refused, although the company later published the logs on the Internet. Kasparov demanded a rematch, but IBM refused and dismantled Deep Blue. Owing to an insufficient sample of games between Deep Blue and officially rated chess players, a chess rating for Deep Blue was not established.

In 2003 a documentary film was made that explored these claims. Entitled Game Over: Kasparov and the Machine, the film implied that Deep Blue’s heavily promoted victory was a plot by IBM to boost its stock value.

One of the two racks that made up Deep Blue is on display at the National Museum of American History in their exhibit about the Information Age; the other rack appears at the Computer History Museum in their « Mastering The Game: A History of Computer Chess » exhibit.

Feng-hsiung Hsu later claimed in his book Behind Deep Blue that he had the rights to use the Deep Blue design to build a bigger machine independently of IBM to take Kasparov’s rematch offer, but Kasparov refused a rematch (see also Hsu’s open letter about the rematch linked below). Kasparov’s side responded that Hsu’s offer was empty and more of a demand than an offer because Hsu had no sponsors, no money, no hardware, no technical team, just some patents and demands that Kasparov commit to putting his formal world title on the line before further negotiations could even begin (with no guarantees as to fair playing conditions or proper qualification matches).

Deep Blue, with its capability of evaluating 200 million positions per second, was the fastest computer that ever faced a world chess champion. Today, in computer chess research and matches of world class players against computers, the focus of play has often shifted to software chess programs, rather than using dedicated chess hardware. Modern chess programs like Rybka, Deep Fritz or Deep Junior are more efficient than the programs during Deep Blue’s era. In a recent match, Deep Fritz vs. world chess champion Vladimir Kramnik in November 2006, the program ran on a personal computer containing two Intel Core 2 Duo CPUs, capable of evaluating only 8 million positions per second, but searching to an average depth of 17 to 18 plies in the middlegame thanks to heuristics.

Kasparov’s loss to Deep Blue inspired the creation of a new game called Arimaa.

[/lang_en]