The early days: when humans thought themselves unmatchable
For decades, chess players — including the greatest champions — were convinced that a machine could never beat the best human player. That conviction owed less to rational analysis than to a form of intellectual pride: chess seemed to embody the very form of intelligence that machines could never reproduce — creativity, intuition, long-term vision.
The first chess programs of the 1960s and 1970s confirmed this scepticism. Programs such as Mac Hack (1967) played mechanically, at the level of an average club player; it would take Belle, in the early 1980s, for a machine to reach master level. Grandmasters dismissed them with contempt.
Deep Blue: the match that changed history
In 1997, IBM unveiled Deep Blue, a supercomputer dedicated to chess, capable of analysing 200 million positions per second. The match against Garry Kasparov, then reigning world champion and world number one for more than ten years, was billed as the ultimate test of artificial intelligence against human intelligence.
- 1996 — Deep Blue I: Kasparov beats Deep Blue 4–2. The machine wins one historic game (the first defeat of a reigning champion by a computer under normal conditions), but Kasparov adapts and wins the match.
- May 1997 — Deep Blue II: Deep Blue, extensively reconfigured, beats Kasparov 3.5–2.5. Game 2 is controversial: Deep Blue plays a move (36.axb5, declining the material gain with 36.Qb6) that Kasparov judges impossible for a 1997 machine. At the time IBM refuses to show the machine’s logs. The controversy continues to this day.
- After 1997: IBM dismantles Deep Blue immediately, refusing a rematch. The story will never be closed in a way that satisfies Kasparov.
The Deep Blue controversy: Murray Campbell, one of Deep Blue’s designers, revealed that at move 44 of game 1 a bug had made the machine play a near-random move, which Kasparov is said to have read as a sign of unsettling depth. The irony is that this “stroke of genius” was an error.
After Deep Blue: the era of commercial engines
In the 2000s, commercial chess engines (Fritz, Junior, Rybka) quickly overtook the level of any human player. Vladimir Kramnik, world champion in 2006, lost a public match to Deep Fritz 4–2. By 2007–2008, the difference in strength between the best humans and the best engines reached several hundred Elo points.
This situation radically transformed opening preparation. From then on, all the grandmasters worked with engines to analyse openings and positions, discover moves that human intuition would never have found, and prepare specific variations against their opponents.
Stockfish: the open-source giant
Stockfish is today the strongest chess engine in the world. Built on alpha-beta search, it has incorporated a neural-network evaluation (NNUE) since 2020. Open source since its creation in 2008, it benefits from the contributions of hundreds of developers around the world. Its estimated Elo exceeds 3600, some 800 points above the best human player.
Stockfish has revolutionised top-level chess practice by allowing every player, from any desktop computer, to analyse their games at a level of quality that IBM achieved in 1997 with millions of dollars of infrastructure. The democratisation of analysis is total.
AlphaZero: the reinforcement learning revolution
In 2017, DeepMind (a Google subsidiary) published the results of AlphaZero, a program of a radically different nature from all its predecessors. Where Stockfish is based on manually programmed rules and encyclopaedic opening tables, AlphaZero is a neural network that learned chess by playing against itself for 9 hours, starting from nothing, with only the rules of the game.
After 9 hours of self-learning (it had already overtaken Stockfish after 4 hours), AlphaZero beat Stockfish 28–0 (with 72 draws) in a 100-game match. This stunning result sent a shockwave through the worlds of chess and AI.
“I can’t disguise my satisfaction that it plays with a very dynamic style, much like my own!” — Garry Kasparov, Science, December 2018
What the AlphaZero style teaches us
The analysis of AlphaZero’s games has profoundly influenced modern opening theory. The program developed positional ideas that humans and classical engines had not found — notably long-term pawn sacrifices to gain the initiative, and an extremely high valuation of piece mobility.
Several ideas judged dubious by classical theory have been rehabilitated by AlphaZero. Its aggressive h-pawn push (h4-h5) in quiet positions, in particular, has been imitated by the best players in the world.
Leela Chess Zero: open-source AI
Leela Chess Zero (Lc0) is the open-source counterpart of AlphaZero. Developed by the open-source community, it reproduces the same neural network architecture and is trained on millions of self-played games. Its strength is now comparable to Stockfish in many positions.
The coexistence of these two types of engine is interesting: Stockfish and Lc0 have different strengths depending on the type of position, and their evaluations sometimes diverge significantly. When the two disagree, it is often in positions of extraordinary complexity.
AI and online cheating
The rise of engines has also created an acute problem for online chess: computer-assisted cheating. On platforms such as Chess.com or Lichess, detection algorithms analyse suspicious games and compare the moves played with those recommended by engines. Detection is statistically robust but never 100% certain, creating recurring controversies in the community.
The Niemann–Carlsen affair of 2022 — in which Magnus Carlsen withdrew from a tournament after losing to Hans Niemann, implying cheating, before Chess.com published a controversial report — is the symbol of these tensions. Cheating remains the unsolved challenge of chess in the digital age.
Play chess in the tradition of the physical game
In an age when engines rule, nothing replaces a real solid wood chessboard and fine Staunton pieces. Physical play is irreplaceable for developing chess intuition and sensitivity.
Choose a wooden chessboard →AI has profoundly transformed opening theory — to understand these modern openings, explore the Sicilian Defence and the Queen’s Gambit as they are analysed in the AlphaZero era. And for the story of the man most marked by this revolution, read the biography of Garry Kasparov.