Deep Blue versus Kasparov

There were two matches, and the second is remembered at the expense of the first, which is a loss, because the first contains the sentence that explains everything that followed.

Philadelphia, February 1996, over six games. Deep Blue won game one on 10 February; The Week In Chess wrote that the match "got off to the ideal start for the sponsors with a Kasparov loss," and the Computer History Museum marks the date for the reason everyone else did: "No computer had ever won a game against a world champion in chess." Kasparov then took the match 4–2, winning the final two games, and said: "I won this match 4 to 2, but it was as tough as a world championship match and, believe me, I played very very seriously." The sentence that matters came after game four, when the match was level: "I'm tired from these games and if I was playing a normal human match my opponent would also be exhausted, but here I have something that is not exhausted and it's playing with the same strength." The asymmetry that decided the rematch was named by its victim fifteen months in advance.

New York, May 1997, for a purse of 1.1 million dollars. The crosstable: Kasparov won game one; Deep Blue won game two; games three, four and five were drawn; Deep Blue won game six. Final score 3.5 to 2.5 — the contemporaneous TWIC report computes a performance rating of 2852 for the machine, against Kasparov's then-rating of 2795.

The match turned on a game Kasparov did not have to lose. In game two he resigned in a position later shown to be holdable, and he never let it go: "I think game two posed too many questions and I think it is in the interest of the IBM team to answer those questions," he said the following week, having already told the closing press conference, "I resigned today. I think the crucial game was game two... I never recovered after this game." Game six, on 11 May, lasted 19 moves; TWIC's verdict was that "strictly speaking game 6 was effectively resigned by Kasparov on move 7," his 7...h6 in a Caro-Kann being the moment the report singles out. The champion who had ground out the 1996 comeback did not show up; the report reads his collapse as the loss of "the mental battle with himself."

Kasparov's demand at the final press conference was not for sympathy: "But I think it's very important for all of us to state today that Deep Blue now must enter competitive chess." It never did. Four weeks later TWIC was already reporting the aftermath as routine: Kasparov's "first appearance since his debacle against Deep Blue" at a tournament in Novgorod, other grandmasters losing exhibition games to programs — matches the editor called "both unedifying yet inevitable" — and FIDE's president promising to sort out the status of computers in tournaments after consulting, of all bodies, the International Olympic Committee.

The machine's side of the story is the plainer one: an evaluation engineered by a team, a database of grandmaster games, and search on special-purpose hardware at a scale no human process resembles. Nothing about the 1997 result said machines understood chess. It said that against enough well-directed computation, that distinction stopped deciding games — which is why this match, and not any later one, is the moment the argument about what "real" chess understanding is went from philosophical to practical, and why the argument has been repeating itself in other domains ever since.