
Co-founder and Chief AGI Scientist, Google DeepMind
A DeepMind co-founder studying general intelligence, its measurement, and the implications of increasingly capable AI.
Shane Legg is a New Zealand-born researcher and co-founder of DeepMind. He studied computer science and completed his doctorate at the University of Lugano through research associated with IDSIA. His early work examined how intelligence could be described mathematically and how an agent's ability to succeed across environments might be evaluated.
With Marcus Hutter, Legg developed work on universal intelligence, seeking a formal measure that brought together ideas from computation and decision making. These theoretical questions differ from the practical measurement of performance on a particular benchmark. They helped frame his continuing interest in systems whose abilities extend across a wide range of tasks.
Legg co-founded DeepMind in 2010 with Demis Hassabis and Mustafa Suleyman. The organization developed research programs spanning reinforcement learning, games, and other applications of machine learning. Its later achievements were the work of substantial collaborative teams, rather than inventions attributable to its founders alone.
He serves as DeepMind's chief artificial general intelligence scientist. His work concerns the possibility of broadly capable AI and the questions that follow from its development, including evaluation and safety. Legg's contribution combines theoretical investigation with laboratory building and research direction. It reflects the continuing challenge of connecting abstract ideas about intelligence with systems whose capabilities can be measured and understood in practice.
Proposed a formal account of intelligence with Marcus Hutter.
Established the AI research company with Demis Hassabis and Mustafa Suleyman.
Investigated mathematical approaches to understanding general machine intelligence.