Shane Legg is a New Zealand-born machine learning researcher who co-founded DeepMind in 2010 with Demis Hassabis and Mustafa Suleyman, and serves as its Chief AGI Scientist. He helped popularize the term “artificial general intelligence” in the early 2000s alongside Ben Goertzel, and his work has long centered on what intelligence is and how to measure it.
Legg earned his PhD in 2008 from the Dalle Molle Institute for Artificial Intelligence (IDSIA), affiliated with the Universita della Svizzera italiana, under the supervision of Marcus Hutter. His thesis, “Machine Super Intelligence,” extended Solomonoff’s model of universal sequence prediction to the active, decision-making case, building on Hutter’s AIXI agent. From this he derived a universal measure of intelligence intended to apply across many kinds of environments rather than a single task. The work was recognized with a research prize from the Singularity Institute.
At DeepMind, Legg has helped steer the lab’s long-run pursuit of general intelligence while also being an early and consistent voice on the risks that powerful AI could pose. His thesis remains a touchstone for attempts to give “intelligence” a precise, formal definition.