Who is Demis Hassabis?
Demis Hassabis was born in London on 27 July 1976. His career connects computer science, neuroscience and artificial intelligence development, including the 2024 Nobel Prize in Chemistry for work on protein structure prediction. [5]
He currently chairs Google DeepMind and serves as Alphabet’s chief scientist, while remaining Isomorphic Labs’ chief executive. [6]
Chess, video games and education
He excelled at chess as a child and helped develop Theme Park at 17, an early stage of his work with interactive systems. [2]
He studied computer science at Queens’ College, Cambridge, from 1994 to 1997. He worked on AI programming for Black & White at Lionhead and founded Elixir Studios in 1998. The studio developed Republic: The Revolution and Evil Genius. In 2005 its rights and technology were sold, and Hassabis returned to academic research. [7]
He completed a cognitive neuroscience doctorate at University College London in 2009, with Eleanor Maguire as primary supervisor and Geraint Rees as secondary supervisor. He continued as a postdoctoral researcher at UCL’s Gatsby unit. This training focused on mechanisms of memory and imagination. [8]
Memory, imagination and learning
In 2007, with Dharshan Kumaran, Seralynne Vann and Maguire, he studied how people with hippocampal amnesia imagined new situations. Four of the five patients struggled to construct coherent scenes. The finding linked memory and imagination through the spatial organisation of experiences, while also showing differences between patients. [9]
In a 2017 review, Hassabis and colleagues advocated closer exchange between neuroscience and AI. They examined how understanding learning and neural computation can suggest new artificial methods. The proposal concerned principles and hypotheses for research rather than reproducing every biological detail of the brain. [10]
Founding DeepMind
In 2010 he founded DeepMind with Shane Legg and Mustafa Suleyman. The venture brought research and entrepreneurship together around developing artificial intelligence systems with more general capabilities. [11]
Google acquired the company in 2014. [8]
Reinforcement learning and simulated environments
He coauthored the DQN paper published in Nature in 2015 with Volodymyr Mnih, Koray Kavukcuoglu, David Silver and other researchers. The agent learned Atari games from images and scores. A common architecture and procedure were evaluated across 49 games, with a different model trained for each one. [12]
DeepMind Lab, introduced in 2016, provided a three-dimensional platform for researching agents in worlds they can observe only partially. Hassabis is among its authors. These environments enable task design, changes to visual conditions and investigation of learning and navigation within a controlled computational experiment. [14]
AlphaGo and AlphaZero
AlphaGo combined neural networks for selecting moves and evaluating positions with search algorithms. After defeating European champion Fan Hui in 2015, it beat Lee Sedol 4–1 in Seoul in March 2016. The system initially learned from human games and then improved by playing against versions of itself. [4]
In 2017 David Silver’s team, with Hassabis as a coauthor, introduced AlphaZero. A common algorithm learned chess, shogi and Go through self-play, starting from each game’s rules. Its generalisation concerned these formalised domains, with defined rules and outcomes. [13]
AlphaFold and the Nobel Prize in Chemistry
AlphaFold2 excelled in the blind CASP14 assessment in 2020; John Jumper and colleagues published the paper in 2021. The system predicts three-dimensional structures from amino acid sequences, incorporating evolutionary information and structural knowledge. It also provides confidence estimates to help interpret its outputs. [15]
In 2022 DeepMind and EMBL-EBI expanded their public database to approximately 200 million protein structure predictions. Open access turned the computed models into a resource researchers across disciplines can consult, using them as a starting point for developing and testing hypotheses. [16]
In 2024 Hassabis and Jumper shared half the Nobel Prize in Chemistry for protein structure prediction. The other half went to David Baker for computational protein design. The prize distinguished two related but different contributions: predicting structures and designing new proteins. [18]
AlphaFold3, published in 2024 by a Google DeepMind and Isomorphic Labs team, extended the approach to complexes involving proteins, DNA, RNA and small molecules. It used a diffusion model to generate atomic positions. The paper describes limitations including stereochemical errors and difficulties representing molecular dynamics. [17]
Isomorphic Labs and drug discovery
In November 2021 he announced Isomorphic Labs as a new Alphabet company. Its aim was to apply AI to drug discovery with a team combining computer science, biology, medicinal chemistry and biophysics. Hassabis took the chief executive role to establish its strategy and facilitate collaboration with DeepMind. [19]
In September 2026 the company described IsoDDE, its drug design engine, and experiments generating and selecting molecules followed by synthesis and laboratory testing. It reported preclinical findings and preparation for clinical development. This sequence distinguishes computational prediction, experimental testing and subsequent evaluation in patients. [20]
Google DeepMind, Gemini and new responsibilities
In April 2023 DeepMind and Google Brain combined as Google DeepMind under Hassabis’s leadership. The reorganisation sought to bring research, computing infrastructure and product development closer together within Google. [21]
In December 2023, he presented Gemini 1.0 on his team’s behalf: models combining text, code, images, audio and video. The project brought together contributions from different Google teams.[22]
In 2026 he handed over daily operational leadership to become Google DeepMind’s chair and Alphabet’s chief scientist. Koray Kavukcuoglu took charge of the unit as senior vice-president. Hassabis retained scientific advisory responsibilities and his leadership of Isomorphic Labs. [6]
Model evaluation and data handling
Initiatives developed under his leadership include the February 2025 update to the Frontier Safety Framework. It relates critical capabilities to safeguards, reviewable safety arguments and deployment decisions. It includes subsequent review and distinguishes misuse threats from potential problems in controlling autonomous systems. [23]
DeepMind’s healthcare work also faced scrutiny. In 2017 the UK data protection authority identified deficiencies in the Royal Free’s agreement with DeepMind and its prior assessment of the processing. The hospital’s corrective undertaking acknowledged improvements already made and required measures concerning purposes, information and data protection. [24]
Honours and accounts of his career
He was elected a Fellow of the Royal Society in 2018. That year he received a CBE, and in 2020 the Dan David Prize in the future category for his AI work. [2]
AlphaFold’s development also received the 2023 Lasker Basic Medical Research Award, Breakthrough Prize in Life Sciences and Gairdner International Award. These honours preceded the Nobel and recognised Hassabis, Jumper and their team’s contribution to structural biology. [3]
In 2024 he received a knighthood for services to artificial intelligence. [27]
The Thinking Game, directed by Greg Kohs and premiered at the 2024 Tribeca Festival, follows his career and DeepMind’s work. It uses archival material and observation of the team to present the development of their scientific projects. [25]
The Infinity Machine, Sebastian Mallaby’s biography of Hassabis and DeepMind, was published in March 2026. The book examines the relationship between scientific ambition, business organisation and artificial intelligence development. [26]
