Much ink, and the digital equivalent, has been spilled on the subject of the second Garry Kasparov-Deep Blue match, won by the machine 3.5-2.5. Many, myself included, believe that Kasparov was still the stronger player at the time; his problem was that he managed to psych himself out by the time of game 6, which he lost horribly. That psychological collapse is believed to have started with his first loss in game two, when Deep Blue made a pair of surprisingly "human" moves: 37.Be4, which was human in a good way (a fine preventive move eschewing material gain to keep Black bottled up and without counterplay), and 44.Kf1, which allegedly blundered into an eventual perpetual check. (Kasparov, trusting the computer, resigned a move later.)
It is clear that the game, and the mysteriousness of those two moves, affected him strongly. But according to statistician Nate Silver, Kasparov's psychological confusion began in game 1, when the computer made a rather silly move due to a bug. Silver thinks (or maybe reports) that Kasparov took this as a sign of the depth of Deep Blue's algorithm rather than as a bug, and the spooking began.
Is it true? Who knows. It's an interesting story, though.
(HT: Ken Regan)