Deep Blue
loreprogrammingops culture
The IBM machine that defeated world chess champion Garry Kasparov in May 1997, the first machine victory over a reigning champion in match play.
Deep Blue won the 1997 rematch 3.5 to 2.5, a year after losing the first match, evaluating around 200 million positions per second with custom chess chips and brute search rather than anything resembling understanding. Kasparov suspected human intervention behind one uncanny move; IBM declined a third match and retired the machine. The victory redrew the line the Turing test had sketched: machines could now beat us at the games we called intelligence.
In May 1997 IBM's Deep Blue beat Garry Kasparov in a six-game match, the first time a computer defeated a reigning world champion under standard conditions. The machine evaluated enormous numbers of positions per second and did not think about chess in anything resembling the way Kasparov did.
The episode is remembered partly for a bug. In the first game Deep Blue made a move that Kasparov read as deep strategy, and which the team later indicated came from a fallback when the machine could not decide. Kasparov's belief that he was facing something more sophisticated than he was appears to have affected the rest of the match, which is a genuinely interesting result about psychology rather than about computation.
What it settled and did not settle is the useful part. It settled that chess is amenable to search plus evaluation at sufficient scale, which many had doubted. It did not settle anything about understanding, and the later systems that learned the game from self-play rather than from encoded human knowledge are a more interesting result. Deep Blue was an engineering achievement aimed at one problem, and the framing of it as a milestone for machine intelligence was mostly supplied by everyone except the engineers.