Don’t miss this insightful chronology of the key moments in the relationship between artificial intelligence (AI) and chess.

Why a timeline about the relationship between chess and AI?

Today, both chess and artificial intelligence (AI) enjoy widespread popularity. You could say that chess and AI are reconnecting, almost like siblings, with chess as the older brother.

If AI is based on the idea of creating programs capable of performing operations comparable to those of an “intelligent” human being, and chess has always been viewed as a game for “intelligent” people, it seems almost natural for a close relationship to exist between the two.

Let’s not question what it means to be “intelligent” for now. Instead, let’s explore the strong bond between AI and chess. But before doing so, we need to look back at the key moments where their paths first crossed.

Brief history of Artificial Intelligence and chess (From Alan Turing to Demis Hassabis)

June 23, 1912 – Alan Turing is born in London, England.

1941 – Turing begins theorizing about machine intelligence, possibly as early as early 1941, when he shares an essay with colleagues at the Government Code and Cypher School. This lost essay is considered the first work in the field of AI. Around this time, Turing also discussed the mechanization of chess with Donald Michie.

1945 – Turing writes his report “Proposed Electronic Calculator”, also known as “‘Proposals for the Development of an Automatic Computing Engine (ACE)”. This design includes brief discussions on computational intelligence and chess, hinting at the boundaries of AI.

‘Can the machine play chess?’ It could fairly easily be made to play a rather bad game. It would be bad because chess requires intelligence. We stated at the beginning of this section that the machine should be treated as entirely without intelligence. There are indications however that it is possible to make the machine display intelligence at the risk of its making occasional serious mistakes. By following up this aspect the machine could probably be made to play very good chess.’ (“Proposed Electronic Calculator”, taken from the book The Essential Turing, page 374)

February 20, 1947 – Turing delivers what is considered the first public lecture mentioning computational intelligence. In it, he discusses the possibility of machines acting intelligently, learning, and even defeating human opponents at chess. He envisioned a machine that could learn from experience and modify its instructions autonomously.

1948 – Turing, along with his colleague D.G. Champernowne, begins writing a chess program for a computer that did not yet exist. The machine’s behavior was simulated manually using pen and paper. During and after World War II, Turing experimented with machine routines for playing chess.

In the same year, Turing publishes a report for the National Physical Laboratory titled “Intelligent Machinery”. According to Australian roboticist Rodney Allen Brooks, this report is even more significant than “Computing Machinery and Intelligence”, where the Turing Test is introduced.

“Intelligent Machinery” already introduces a precursor to the Turing Test and provides examples of how simple computational mechanisms can adapt, be trained, and learn independently.

November 1951 – Dietrich Prinz writes the first chess program to run electronically on the computer at the University of Manchester.

1953 – A year before his death (June 7, 1954), Turing publishes his essay “Chess”, which focuses on computer chess. This represents his final published work related to AI.

In this essay, Turing anticipates many foundational concepts of computer chess, such as using heuristics to guide the tree of possible moves, evaluation rules to assign numerical values to strengths and weaknesses, and the minimax strategy.

Additionally, the learning procedure Turing proposes involves the machine testing variations in its playing method and adopting those that lead to better results. This approach is an early example of what is now known as a genetic algorithm (GA), a term introduced by John Holland around 1975.

1965 - In 1997, John McCarthy wrote that, back in 1965, the Russian mathematician Alexander Kronrod coined the famous phrase: “Chess is the Drosophila of Artificial Intelligence.”

The Drosophila, commonly known as the fruit fly, has been widely used in genetic research. Early pioneers of chromosome theory studied fruit flies instead of larger animals like chickens or cows to develop genetic mapping methodologies. Fruit flies are convenient for research because they are easy to breed and share 75% of the genes that cause diseases in humans.

This analogy suggests that chess, better than other open-ended intellectual tasks, serves as a perfect model for studying how human knowledge can be represented in machines.

November 22, 1967 - The first chess match between two computers takes place. The game lasted nearly nine months. A chess program developed by Georgi Adelson-Velskiy and Alexander Kronrod at Moscow’s Institute of Theoretical and Experimental Physics (ITEP) competed against the Kotok-McCarthy program running at Stanford. The result was a 3-1 victory for the Russian program.

1968 - British AI researcher Donald Michie and American computer scientist John McCarthy bet Scottish International Master David Levy that a computer would defeat him within 10 years. Levy won the bet.

July 27, 1976 - Demis Hassabis is born in the United Kingdom to a Greek father and a Singaporean mother. At the age of 4 (in 1980), he begins playing chess.

1983 - Ken Thompson’s chess machine, Belle, achieves the level of a chess master.

1987 - According to Hassabis in Game Changer, at age 11, after losing in an international chess tournament in Liechtenstein, he began pondering what could be achieved if all the brilliant minds involved in games like chess focused their efforts on fields like science or medicine.

1989 - At just 13 years old, Hassabis becomes a chess master. He is ranked second in the world among players under 14, behind only Judit Polgar from Hungary.

In the same year, IBM launches the Deep Blue project. Feng-Hsiung Hsu and Murray Campbell join IBM Research to work on parallel processing systems. Feng-Hsiung Hsu develops the chess program Deep Thought.

February 1996 - Garry Kasparov defeats IBM’s Deep Blue, though the computer manages to win one game.

May 11, 1997 - An upgraded version, Deeper Blue, defeats world champion Garry Kasparov in a six-game match. Deep Blue relied on traditional game-playing algorithms and brute-force computation, exhaustively exploring all possible moves to find the optimal solution.

Although this victory is often seen as a triumph for AI, IBM saw it differently. Deep Blue didn’t replicate human-like thought processes, and there was no formula for intuition.

Computer scientist Nils John Nilsson, however, argued that Deep Blue demonstrated the strength of chess programs, which could still benefit from integrating human knowledge, skills, and machine learning techniques.

This event marked the fulfillment of a long-standing prediction that a computer would one day beat a high-level player. Turing, Shannon, and others had speculated about this as early as the 1940s and 1950s.

2006 - The last major chess match between a human and a computer takes place. World champion Vladimir Kramnik loses 4-2 to Deep Fritz.

2010 - Demis Hassabis and Shane Legg found DeepMind Technologies, an artificial intelligence research company.

2016 - AlphaGo, DeepMind’s AI system, defeats the renowned South Korean Go champion Lee Sedol. The creative and unconventional ways in which AlphaGo played were seen as groundbreaking for AI.

2017 - Building on AlphaGo’s success, DeepMind begins its most ambitious project yet: AlphaZero. The key concept is generality—creating a single system capable of excelling in various tasks. AlphaZero is designed to play chess, shogi, and Go.

December 17, 2017 - A team of 17 researchers at DeepMind publishes a paper introducing AlphaZero, a self-learning system capable of mastering chess, shogi, and Go from scratch.

AlphaZero outperformed other champion programs. In chess, it defeated Stockfish 8. Out of 100 games, AlphaZero won 28, while the remaining 72 ended in draws. In a broader test of 1,000 games, AlphaZero claimed 155 victories compared to Stockfish’s 6.

Traditional programs like Stockfish and Deep Blue relied on countless rules and heuristics provided by top players, designed to cover all possible contingencies. AlphaZero, however, used neural networks and general-purpose algorithms, requiring only the basic rules of the game to achieve mastery.

This timeline highlights the pivotal role chess has played in the development of artificial intelligence. Today, chess reaps the benefits of this historic relationship, as innovations like AlphaZero have introduced groundbreaking strategies and techniques previously dismissed as impractical or nonsensical.

For further reading, we recommend the following books and texts, which were key sources for this timeline. But first, we also recommend you to read our article about chess 960 , where we also make an approach to the relationship between artificial intelligence and chess.

Read more

B. Jack Copeland’s The Essential Turing….

Nils J. Nilson’s The Quest for Artificial Intelligence (2010).

The famous book of Matthew Sadler and Natasha Regan, Game Changer (2019).

S. Barry Cooper’s and J. van Leeuwen’s Alan Turing: His Work and Impact (2013).

AlphaZero shedding new light on chess and Go

Demis Hassabis: 15 facts