The matches between IBM's chess computer Deep Blue and reigning world chess champion Garry Kasparov, played in 1996 and 1997, were a landmark moment in the public perception of machine intelligence. The 1997 rematch, in which Deep Blue defeated Kasparov, marked the first time a computer had beaten a reigning world champion in a multi-game match under standard tournament time controls.
1996 match
The first match, held in Philadelphia in February 1996, was won by Kasparov four games to two, including one loss to Deep Blue in the opening game, the first time a computer had beaten a world champion in a single game played under normal time controls. Despite the loss of that first game, Kasparov adjusted his strategy and won the overall match convincingly, and the result was widely reported as a reassuring demonstration that human intuition and adaptability still outmatched brute-force computer search.
1997 rematch
IBM substantially upgraded the system for a rematch held in New York in May 1997, roughly doubling its search capacity to evaluate on the order of 200 million chess positions per second. Deep Blue won the six-game rematch 3.5 to 2.5, including a decisive final-game win in under an hour that Kasparov later said unsettled him because of a move he could not explain through conventional search-based play, fueling his suspicion, never substantiated, that IBM engineers had intervened during the game. Kasparov requested a further rematch, which IBM declined, retiring Deep Blue from competitive play shortly afterward.
Significance
The 1997 defeat became one of the most widely cited early symbols of machine capability exceeding elite human performance in a narrow, well-defined domain, predating by nearly two decades similar milestones in other games such as AlphaGo's 2016 victory over Lee Sedol at the game of Go, an event with far greater branching complexity than chess. Deep Blue's approach, relying on specialized hardware and brute-force search through vast numbers of positions guided by hand-crafted evaluation functions, rather than learned strategy, is frequently contrasted with later systems like AlphaZero, which learned chess, shogi, and Go from self-play without hand-coded chess knowledge. IBM later pursued a different AI showcase with Watson's 2011 Jeopardy! victory, and the Deep Blue matches are sometimes discussed alongside the Turing test as an early benchmark framing for machine intelligence claims, commonly cited as an early illustration of the gap between narrow, task-specific machine competence and broader general intelligence.