How did a computer beat the world chess champion?

Mostly by brute speed, plus a great deal of human chess knowledge built in. In May 1997 IBM's Deep Blue beat Garry Kasparov 3½–2½ in a six-game match in New York, the first time a reigning world champion lost to a computer under tournament conditions. The machine evaluated 200 million positions per second and usually looked six to eight moves ahead, sometimes twenty or more. Its rules for judging positions were written and tuned by chess masters and programmers, and it carried a library of 700,000 grandmaster games.

Kasparov had won their first match a year earlier, 4–2. For the rematch the hardware was upgraded to twice the speed, and the match was also fought in the champion's head. In the first game, a bug made Deep Blue choose a random legal move. Kasparov, not knowing this, read the strange move as a sign of superior intelligence. After losing the second game, he accused IBM of having a grandmaster secretly help the machine; IBM denied it, and analysis later showed Kasparov could have saved that game with a draw.

The decisive sixth game was over in under twenty moves. Kasparov chose a dubious opening, betting that a computer would never sacrifice a knight without an immediate gain. Deep Blue did exactly that, and his defence collapsed.

The win was celebrated as a sign that artificial intelligence was catching up with humans. Kasparov, who once called Deep Blue an "alien opponent", later dismissed it as "as intelligent as your alarm clock".