In the 1990s, it was clear that computer chess was reaching an important inflection point. The top computer chess programs were now capable of beating even the best players in the world in blitz games, and these same programs were at least competitive against very strong human opposition, even if they had not proven that they were yet ready to topple elite grandmasters under tournament conditions.
In 1996, IBM attempted to change this through the creation of a supercomputer known as Deep Blue. This chess-playing computer was created with one purpose in mind: defeating the human chess World Champion in a match played under conditions as close to those found in a competitive match or tournament situation as possible.
At the time, the World Chess Champion was none other than Garry Kasparov, considered by many to be the greatest chess player ever. The first match between the two was played in February 1996, and a second match was later played in May 1997.
About the Computer
Deep Blue was a processing monster. Most of the computer's strength was derived from the massive supercomputer itself, which was, by the time of the 1997 match, ranked among the 300 most powerful supercomputers in the world. Deep Blue was capable of evaluating around 200 million positions each second, allowing searches typically in the range of 13-15 ply per move.
In addition, several grandmasters were involved in the evolution of Deep Blue's evaluation functions. Miguel Illescas, John Fedorowicz and Nick de Firmian worked on the opening book, while Joel Benjamin helped Deep Blue increase its "knowledge" of chess by improving its evaluation parameters. Finally, tablebases were used to allow the computer access to flawless play in five-piece endgames, as well as some six-piece endgames.
The First Match
The first match between Kasparov and Deep Blue was scheduled for February 1996. The six-game match was played in Philadelphia, Pennsylvania.
Heading into the match, Kasparov was seen as the favorite, in large part because a computer had never defeated a World Chess Champion in a formal setting at normal time controls. However, some opinions were changed after the first game, when Deep Blue shocked the world by winning with the white pieces in 37 moves.
However, Kasparov quickly recovered. In the second game, Kasparov played a Catalan and prevented his computer opponent from fully developing its pieces in the early going. After 73 moves, the operators of Deep Blue resigned in a hopeless endgame.
The match was now tied at 1-1, and it was clear that both sides were capable of scoring against the other. In the next two games, Kasparov and Deep Blue played fairly equally, with both games ending in draws to take the match to 2-2 with two games to play.
Game Five would provide to be a critical turning point in the match. Kasparov switched away from playing the Sicilian Defense, instead choosing to go for a Four Knights Game. Deep Blue had every chance to escape from the game with a draw, but its team chose to decline a draw offer from Kasparov on the 23rd move. Unfortunately for the machine, things went badly after the offer, and Kasparov claimed victory after 47 moves - the only win for Black in the match.
This meant that Kasparov needed only a draw to win the match, while Deep Blue would require a win in the final game to claim a 3-3 tie. Kasparov, however, won the last game in convincing fashion, thoroughly outplaying the machine to make the final margin of victory 4-2 in his favor.
The Second Match
After IBM spent the next year upgrading Deep Blue, they decided they were ready to challenge Kasparov once again. The second match began on May 3, 1997, and started in auspicious fashion for the machine, as Kasparov took the white pieces and won in 45 moves to take an early 1-0 lead.
However, things began to change in the second game - though the manner in which the match proceeded from this point on was not without controversy. Deep Blue won the second game when Kasparov resigned after 45 moves. However, later analysis showed that Kasparov had resigned in a drawn position; Kasparov could have forced a perpetual check due to a mistake by Deep Blue on the 44th move. It seems as though Kasparov assumed a computer would see such a clearly tactical error, and thus spent little time searching for a saving sequences of checks. This oversight by Deep Blue would later lead Kasparov to believe, controversially, that IBM may have cheated during the match.
In the third game, Kasparov played 1. d3 in an attempt to get the computer out of book, but was unable to gain much of an advantage, and the game was drawn. The fourth game saw Kasparov gain the advantage with Black in a Caro-Kann Defense; however, he was in time trouble and wasted any edge he might have had, bringing the game to a draw once more.
Once again, the match was 2-2 with two games to go, with both sides having reason to believe that could win the match. Kasparov appeared to gain the advantage, but Deep Blue played brilliantly in the endgame to earn a draw and keep the match level at 2.5-2.5.
Game Six was a pivotal game, and one that remains controversial to this day. In the game itself, Kasparov played badly - in an objective sense - in the Caro-Kann Defense, allowing Deep Blue to play a well-known knight sacrifice that ruined Black's position. Kasparov would ultimately resign in just 19 moves.
However, Kasparov found the play of Deep Blue during Game Six quite suspicious, and has repeatedly made the claim that he believes IBM's team cheated in the match. Kasparov did not believe that the computer would have been able to play the strong knight sacrifice found by Deep Blue on its own, and did not expect it to be found in the (by modern standards) relatively small opening book used by the computer.
In 2009, Miguel Illescas gave an explanation for what happened that either shows just how lucky IBM was, or - from Kasparov's perspective - shines even more suspicion on the result of the match. According to Illescas, the Deep Blue team has made some final updates to the Deep Blue opening book the morning of the fateful game. Those updates included adding the knight sacrifice line into the computer's book just hours before the game was played.
For Kasparov, this explanation was less than satisfactory, especially considering that in both games Deep Blue had won, Kasparov played lines he had never before played in a tournament game - yet the opening book was prepared in such a way as to anticipate those lines. While most would point to these as mere coincidences, Kasparov maintains that he was cheated.
While Kasparov desired a rematch, IBM declined, and Deep Blue was soon dismantled. It is a shame that the rematch never occurred; it would have received tremendous publicity, and many experts would likely have expected Kasparov to triumph, considering the unlikely manner in which Deep Blue managed to win the two games it took in the previous match.
While the general public may have taken the match result to mean that computers were now superior to humans at chess, this was unlikely to have been accurate at the time of the match. In fact, top human players managed to win and draw matches against the best computer programs beyond the year 2000; however, by the later years of the first decade of the 21st century (and in particular, after impressive wins by Hydra and Deep Fritz in 2005 and 2006), it was clear that computers had surpassed humans in general chess strength.