Demis Hassabis

• 1. Introduction

  ›

  Demis Hassabis was born Dimitrios Hassapis on the 27th of July 1976, and his very surname carries a story. The original name, Hassapis, means “butcher” in Greek. At some point he executed what one colleague described as “a point mutation” - changing a single letter, p to b, to arrive at the name the world now knows. It is a small act that reflects something larger: a mind that sees the logic hidden inside a system and knows exactly where to intervene.

  His father is a Greek Cypriot who gave up office work to sell toys from the back of a van and chase a career as a singer-songwriter. His mother, a Chinese Singaporean, grew up an orphan, was adopted by a relative, studied nursing, and worked as a retail clerk and part-time cleaner. Their son would go on to win the Nobel Prize in Chemistry in 2024.

  This is a documentary about a man who learned chess by watching his father play against his uncle, taught himself to program from books, designed a bestselling video game at seventeen, earned a PhD in cognitive neuroscience, co-founded one of the most consequential AI laboratories on earth, and cracked a problem that had stumped biologists for fifty years. The question worth asking is not what Hassabis has done. It is how a single mind came to stand at the intersection of so many different fields at exactly the right moment.

• 2. A Prodigy’s Education

  ›

  Chess came first. Hassabis picked up the game at the age of four by watching his father play against his uncle, and by thirteen he had reached master standard with an Elo rating of 2300, captaining England junior chess teams along the way. He would later represent the University of Cambridge in the Oxford-Cambridge varsity chess matches of 1995, 1996, and 1997, winning a half blue.

  His first computer arrived in 1984, a ZX Spectrum 48K bought with chess winnings. He taught himself to program from books, and before long he had written his first AI program on a Commodore Amiga - a reversi-playing opponent he built himself.

  Between 1988 and 1990, he attended Queen Elizabeth’s School in Barnet, a boys’ grammar school in North London. After a period of home-schooling by his parents, he moved to the comprehensive school of Christ’s College in East Finchley and completed his A-level exams two years early, at the age of sixteen.

  When Cambridge University asked him to defer entry because of his young age, the gap year he took would turn out to be anything but idle. He entered an Amiga Power “Win-a-job-at-Bullfrog” competition, and won. At seventeen, working alongside the game designer Peter Molyneux at Bullfrog Productions, he co-designed and served as lead programmer on the 1994 simulation game Theme Park. That game sold several million copies and inspired an entire genre of simulation sandbox games. Despite being offered a seven-figure sum to stay in the games industry, he turned it down. He earned enough from that one year to pay his own way through university.

• 3. Games, Studios, and Early Ambition

  ›

  After graduating from Queens’ College, Cambridge in 1997 with a double first in Computer Science, Hassabis went to work at Lionhead Studios, the company Peter Molyneux had recently founded. There he served as lead AI programmer on the 2001 god game Black & White.

  In 1998 he left to found Elixir Studios, his own London-based independent developer. Publishing deals followed with Eidos Interactive, Vivendi Universal, and Microsoft. Hassabis served as executive designer on two games, Republic: The Revolution and Evil Genius, both of which earned BAFTA nominations for their interactive music scores, composed by James Hannigan.

  Republic: The Revolution was a highly ambitious political simulation that attempted to model the workings of an entire fictional country in AI. Its scope caused serious delays, and the final release was a reduced version of the original vision. Critics were lukewarm; it received a Metacritic score of 62 out of 100. Evil Genius, a tongue-in-cheek parody in the spirit of Austin Powers, fared better at 75 out of 100. In April 2005, Elixir’s intellectual property and technology rights were sold to various publishers and the studio closed.

  The failure of that first ambitious vision did not turn Hassabis away from ambitious visions. It pointed him somewhere else entirely: toward the actual source of intelligence, the human brain.

• 4. Memory, Imagination, and the Brain

  ›

  Eleanor Maguire at UCL’s Queen Square Institute of Neurology supervised Hassabis through his PhD in cognitive neuroscience, which he completed in 2009. He had returned to academia specifically to look inside the human brain for new ideas about how AI might be built differently.

  His very first academic paper, published in PNAS, became a landmark finding. It showed systematically, for the first time, that patients with damage to the hippocampus - the brain region known to cause amnesia - were also unable to imagine themselves in new experiences. The same structure that stores the past also constructs the future.

  From this and a follow-up study using functional magnetic resonance imaging, Hassabis developed a new theoretical account of the episodic memory system. He identified what he called “scene construction” - the process of generating and maintaining a coherent and complex scene in the mind - as the key mechanism underlying both memory recall and imagination. The journal Science listed this work among the top ten scientific breakthroughs of 2007.

  He continued his research as a visiting scientist at MIT, in the lab of Tomaso Poggio, and at Harvard University, before earning a Henry Wellcome postdoctoral research fellowship to the Gatsby Computational Neuroscience Unit at UCL in 2009, working with Peter Dayan. He published work in Nature, Science, Neuron, and PNAS, and eventually extended these ideas into what he called a “simulation engine of the mind” - a system whose function is to imagine future events and scenarios to support better planning. That concept would travel with him into everything he built next.

• 5. Building DeepMind

  ›

  DeepMind was founded in London in 2010 by Hassabis alongside Shane Legg and Mustafa Suleyman. Hassabis had met Legg when both were postdocs at the Gatsby Computational Neuroscience Unit. Suleyman had been a friend through family. Hassabis also recruited his university friend and former Elixir partner David Silver.

  The mission the founders set for themselves was blunt: “solve intelligence” and then use that solution “to solve everything else.” More specifically, the company aimed to combine insights from systems neuroscience with advances in machine learning and computing hardware, with the long-run goal of creating an artificial general intelligence.

  In December 2013, DeepMind announced a significant early result: an algorithm called a Deep Q-Network had been trained to play Atari games at a superhuman level using nothing but the raw pixels on the screen as inputs. The announcement signalled a new approach, one that would come to be called deep reinforcement learning, a field DeepMind helped pioneer by combining deep learning with reinforcement learning.

  In 2014, Google purchased DeepMind for 400 million pounds. The company retained its base in London and operated largely as an independent entity, though DeepMind Health was later incorporated directly into Google Health.

  The purchase gave DeepMind resources to tackle harder problems. Hassabis predicted that artificial intelligence would be “one of the most beneficial technologies of mankind ever” while also warning that significant ethical issues remained unresolved. In 2023, he signed a statement declaring that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

• 6. AlphaGo and the Go Match

  ›

  Go had long been considered a holy grail for AI researchers. Its sheer number of possible board positions, and its resistance to the programming techniques that had conquered chess, made it seem beyond reach for machines. DeepMind set out to change that with AlphaGo.

  In October 2015, AlphaGo beat European champion Fan Hui 5-0. In March 2016, it won 4-1 against former world champion Lee Sedol in what became a famous $1 million challenge match held in Seoul, South Korea. In 2017, it won 3-0 against the world’s top-ranked player, Ke Jie. The Korea Baduk Association awarded AlphaGo an honorary 9-dan Go rank in 2016 - the highest designation in the game.

  The 2016 match was the subject of a documentary film, also titled AlphaGo, which won the Cannes Lion Grand Prix in 2016. Hassabis himself became the subject of a later documentary, The Thinking Game, which premiered at the 2024 Tribeca Festival and was made by the same filmmaker.

  Beyond Go, DeepMind reduced the energy consumed by the cooling systems in Google’s data centres by 40 percent, created a neural Turing machine, and advanced research on AI safety. The company produced nine front cover articles in the journal Nature between 2015 and 2026, and one in Science in 2017.

• 7. AlphaFold and the Nobel Prize

  ›

  Protein structure prediction had been a grand challenge in biology for fifty years. A protein’s function is closely tied to its three-dimensional shape, and working out that shape from just the one-dimensional sequence of amino acids that makes up the protein had defeated researchers for decades. Knowing a protein’s structure can accelerate drug discovery and deepen understanding of disease.

  DeepMind turned its attention to the problem in 2016. At the 13th Critical Assessment of Techniques for Protein Structure Prediction, known as CASP13, in December 2018, AlphaFold won the competition by successfully predicting the most accurate structure for 25 out of 43 proteins. Hassabis described it to The Guardian as “a lighthouse project, our first major investment in terms of people and resources into a fundamental, very important, real-world scientific problem.”

  In November 2020, AlphaFold 2 achieved a median global distance test score of 87.0 across protein targets in the free-modelling category of CASP14 - well above the previous results and with an overall error of less than the width of an atom, below one Angstrom. The organisers of CASP declared the problem essentially solved. AlphaFold v2 was named the Breakthrough of the Year winner by the journal Science in 2021.

  Over the following year, DeepMind used AlphaFold 2 to predict the structures of all 200 million proteins known to science, and made both the system and those structures freely available through the AlphaFold Protein Structure Database, developed in collaboration with EMBL-EBI.

  In 2024, Hassabis and John M. Jumper were jointly awarded the Nobel Prize in Chemistry for their contributions to protein structure prediction. That same year, Hassabis was knighted for his work on AI. In 2026, the biography The Infinity Machine, written by Sebastian Mallaby, was published, adding another kind of structure to the record of his life.

• 8. The Game Player

  ›

  Alongside everything else, Hassabis has remained a serious competitive games player across a remarkable range of disciplines. He is a five-time winner of the World Pentamind Championship at the London Mind Sports Olympiad, taking the title in 1998, 1999, 2000, 2001, and 2003. The Pentamind measures performance across multiple board games simultaneously. He also won the World Decamentathlon Championship twice, in 2003 and 2004.

  In Diplomacy, he was World Team Champion in 2004 and finished 4th at the 2006 World Championship. At the poker table, he has cashed at the World Series of Poker six times, including in the Main Event.

  In 2026, the Korea Baduk Association awarded him a 7th amateur dan in Go, in honour of the tenth anniversary of the AlphaGo versus Lee Sedol match - a personal rank reflecting the game that his own creation had already surpassed at the highest professional level.