Demis Hassabis

Hassabis’s route into AI crossed commercial game development and academic neuroscience. His CV lists work on Theme Park at Bullfrog and the creature-learning systems in Black & White at Lionhead. He founded Elixir Studios in 1998 and ran it until 2005. Along the way, he completed a computer science degree at Cambridge, then returned to research for a cognitive neuroscience PhD at UCL. His doctoral work with Eleanor Maguire investigated how memory and imagination depend on the brain’s ability to construct scenes.[4]

Building DeepMind, then changing roles

Hassabis co-founded DeepMind in 2010. Google acquired it in 2014, and in 2023 combined it with the Google Brain team under his leadership. On 5 August 2026, Google announced that he would hand over day-to-day operating responsibilities and become DeepMind chair and Alphabet chief scientist. Koray Kavukcuoglu took responsibility for leading DeepMind as SVP. Hassabis’s new remit emphasizes AGI strategy and scientific discovery; the transition also gives him more time for Isomorphic Labs, the AI drug-development company he founded.[4][1][3]

A public forum for the AGI debate

With Shane Legg and James Manyika, Hassabis became a director of the DeepMind Institute, launched in September 2026. Its introductory essay argues that decisions about advanced AI require contributions from the humanities, governments and wider society as well as technical researchers. The institute publishes essays and debate about possible AGI outcomes. Its articles carry an explicit distinction between authors’ views and Google’s official position, which matters when reading its forecasts or policy proposals.[11]

AlphaGo combined neural networks that assess positions and propose moves with a search procedure that explores possible continuations. Hassabis is a co-author of the 2016 research paper, which reported a match victory over a professional Go player. AlphaFold addressed a different problem: predicting the three-dimensional structure a protein adopts from information about its amino-acid sequence. The AlphaFold2 paper credits Jumper and Hassabis with leading the research. Their Nobel recognition concerns protein-structure prediction, while David Baker received the other half of the prize for computational protein design.[7][8][2]

Hassabis’s 2024 Nobel lecture connects the development of general learning systems with their use in scientific discovery. It is a useful first watch because it explains the research program in his own words before the technical papers. In a separate July 2026 essay, he proposed a frontier-AI standards body that would develop and update capability tests, work with government agencies, and encourage security and safety practices. The proposed body would coordinate independent testing of increasingly capable AI systems.[9][10]