Shane Legg

Shane Legg is a New Zealand-born computer scientist who co-founded DeepMind in London in 2010 with Demis Hassabis and Mustafa Suleyman, after he and Hassabis met as researchers at UCL's Gatsby Computational Neuroscience Unit. Earlier he had worked as a software engineer at Webmind, Ben Goertzel's AI start-up. In 2002, when Goertzel and Cassio Pennachin were stuck for a title for a book on human-level AI, Legg suggested "Artificial General Intelligence," the name they adopted; Goertzel later learned that Mark Gubrud had used the phrase in 1997.

Legg took a PhD at the Swiss AI lab IDSIA under Marcus Hutter, finishing in 2008 with a thesis titled Machine Super Intelligence. In 2007 he and Hutter gathered some 70 published definitions of intelligence, and in "Universal Intelligence" (December 2007) they proposed a formal measure of it, an agent's ability to achieve goals across all computable environments, weighted toward simpler ones.

At DeepMind he was Chief Scientist and started its AGI safety work, and he was an author, with OpenAI researchers, of the 2017 paper that introduced learning rewards from human preferences, the method later called RLHF. After the April 2023 merger with Google Brain he became Chief AGI Scientist; TIME reported that year that he led its AGI technical safety team. In November 2023 he co-wrote "Levels of AGI," a Google DeepMind framework for grading progress toward general intelligence.

TIME reported in August 2026 that after Hassabis handed daily control to Koray Kavukcuoglu, Legg would keep reporting to Hassabis. On 16 September 2026 Legg, Hassabis and James Manyika launched the DeepMind Institute, a Google publishing platform on AGI's effects on society, with Legg as managing editor.