Chess in the age of AI

Humans compete with machines for second place

In 1968, chess International Master David Levy made a bet with computer scientists that no chess computer would beat him within 10 years. In 1978, he won the bet but lost one game in the match – making him the first Master to lose a match to a computer. Forty-one years later, chess computers (now called chess engines) have taken on a surprising role as human chess players compete for second place. 

A short history of chess computers

Chess computers made a false start in 1770 with a chess machine called the Mechanical Turk, which turned out to be a hoax. The first paper on how computers could be used to play chess was published 180 years later, by American mathematician Claude Shannon.

Computer scientists started working on algorithms based on Shannon’s paper, and in 1957 developed the first chess engines that could play a full game of chess.  Over the next 47 years, computer hardware became more powerful and algorithms improved. 

Finally, in 1997, Deep Blue put an end to human chess dominance by beating the reigning World Chess Champion, Gary Kasparov.


Beyond Deep Blue


Kasparov accused IBM of cheating and requested a rematch; IBM refused and Deep Blue was retired to the National Museum of American History. Deep Blue’s victory forever changed how chess is played. Kasparov became one of the first Grandmasters to train on chess computers to prepare for matches against human opponents. 

How chess engines are changing the game 

Since Deep Blue, chess engines have improved to a point where even the best players in the world stand effectively zero chance of winning a match against a modern chess engine running on a laptop. 

Human players are no longer using chess engines as opponents but as training tools, to analyse games and find better strategies that human players would not have considered. This has led to a significant improvement in the level of chess played by humans. 

Big data and chess


A second line of improvement in chess engines came from big data. The database contains more than 700,000 high-level chess games. These games are analysed to decide opening strategies (called ‘opening books’) and endgame strategies. 

A modern chess engine armed with this database plays chess at a level that cannot be approached by humans. Human players use the statistical data from the opening books to prepare for matches and analyse games played by their opponents. 


A new world chess champion 

In a new tournament called the Top Chess Engine Championship, chess engines compete against each other over thousands of games. The winner of this year’s tournament was an open-source engine called Stockfish that uses a traditional algorithm to evaluate moves. 

Stockfish will have a hard time in the next competition as new-generation chess engines use neural nets and artificial intelligence. The most promising of these engines is Alpha Zero, which uses zero human input or data from human games and learned chess from scratch by playing against itself. It is currently considered the strongest chess engine in the world. 


The future of chess

Chess is an enormously complex game. It is estimated that there are more possible chess games than atoms in the universe. This means not even the best chess engine plays a perfect chess game. Chess engines will continue to evolve and be used in broader applications to help humans solve complex problems. Garry Kasparov predicts that AI engines and humans will eventually work together in all fields to create smarter tools and explore techniques that were never previously considered by humans.