Module 1 · Section 2 of 3

When a Machine Beat the World Champion

Picture the scene: Garry Kasparov, the world’s best chess player, sits across from a refrigerator-sized IBM computer called Deep Blue. The year is 1997. The venue is New York. The match is the rematch — because Deep Blue had lost to Kasparov the year before, and IBM had spent the intervening twelve months making it significantly stronger.

This was not a novelty event. Kasparov was, by any measure, the greatest chess player alive. He had been world champion since 1985. He had crushed every human challenger who came at him. He had played a version of Deep Blue in 1996 and won the match 4–2, losing two games but never looking seriously threatened overall. He came into the 1997 rematch confident.

Game 1, Kasparov won. The script seemed to be repeating itself.

Then something shifted.

In Game 2, Deep Blue made a move that threw Kasparov. Not because it was flashy — because it wasn’t. It was subtle. A quiet repositioning that seemed to sacrifice short-term advantage for something harder to calculate. Kasparov, whose genius was built partly on reading his opponents psychologically, couldn’t figure out what the machine was doing. He eventually resigned a position that, as analysts later showed, he could have drawn.

He was visibly rattled. In the press conference afterwards, he accused IBM of human interference — suggesting that grandmasters must have been coaching the machine mid-game. IBM denied it. The match continued.

Kasparov drew Games 3, 4, and 5. Then, in Game 6, under pressure, he collapsed. He resigned after just 19 moves — one of the shortest defeats of his career. Deep Blue won the match 3.5 to 2.5.

The headlines called it the end of human chess supremacy. Some went further and declared it a turning point in the history of intelligence itself. The reaction was more emotional than analytical: if a machine could beat the best human mind at chess — a game that had been the benchmark of strategic thinking for centuries — what couldn’t machines do?

That reaction was understandable. It was also, as we’ll see, somewhat misplaced.

Why chess seemed to matter so much

For most of the twentieth century, chess was the standard test case for machine intelligence. The logic seemed reasonable: chess requires planning, strategy, and the ability to think several moves ahead. It demands pattern recognition across a vast search space. It punishes short-term thinking. If a machine could master it, the argument went, it would have to be doing something like thinking.

This assumption turned out to be wrong — but not in the way most people expected.

What Kasparov was actually doing

Kasparov didn’t win chess games by calculating every possible move sequence. No human can. The number of possible chess games is larger than the number of atoms in the observable universe. Human players don’t search exhaustively — they prune. They look at a position and their brain immediately filters it down to a handful of candidate moves worth considering. The rest get discarded before conscious analysis even begins.

That filtering is built on years of pattern recognition. A grandmaster looks at a board and instantly recognizes formations — “this pawn structure is weak,” “the king is exposed here,” “this knight will dominate the endgame.” The calculation that follows is deep but narrow. Kasparov’s edge over other humans wasn’t that he could see further ahead than everyone else. It was that he was better at identifying which lines were worth exploring in the first place.

He also read opponents. He knew when someone was under time pressure. He knew which players would panic in certain positions and which would dig in. He used psychology as a weapon. Against another human, that mattered enormously.

Against Deep Blue, it meant nothing.

The moment the frame shifted

The 1997 match forced a reframing that the chess world had resisted for decades: maybe chess wasn’t a test of intelligence in the way everyone assumed. Maybe chess, at its core, was a very hard search problem — and search problems are exactly what computers are built to solve.

This didn’t diminish Kasparov. It clarified what he was doing. Human chess mastery is a genuine cognitive achievement. It just turns out that the cognitive achievement and the computational achievement are solving the same problem through completely different means — and for this particular problem, at this particular moment in hardware development, the computational approach won.

That distinction — same outcome, different mechanism — is the key to understanding everything that comes after.