1973–, Computer scientist; co-founder of DeepMind
Shane Legg is a New Zealand-born computer scientist who co-founded DeepMind with Demis Hassabis and Mustafa Suleyman in 2010. His 2008 PhD thesis at Lugano (under Marcus Hutter) on machine super-intelligence developed an early formalisation of universal intelligence as the expected reward an agent achieves across all computable environments, a quantitative version of Solomonoff's framework adapted for agents.
Legg has remained at DeepMind, where he co-leads the AGI Safety team. He has been a long-standing voice on the importance of AI safety research, predicting in his thesis that human-level machine intelligence would arrive somewhere between 2018 and 2036.
Video
Related people: Demis Hassabis, Mustafa Suleyman
Works cited in this book:
- Human-level control through deep reinforcement learning (2015) (with Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Marc G. Bellemare, Alex Graves, Martin Riedmiller, Andreas K. Fidjeland, Georg Ostrovski, Stig Petersen, Charles Beattie, Amir Sadik, Ioannis Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Demis Hassabis)
- Deep reinforcement learning from human preferences (2017) (with Paul Christiano, Jan Leike, Tom B. Brown, Miljan Martic, Dario Amodei)
- Scalable agent alignment via reward modeling: a research direction (2018) (with Jan Leike, David Krueger, Tom Everitt, Miljan Martic, Vishal Maini)
- Avoiding Side Effects By Considering Future Tasks (2020) (with Victoria Krakovna, Laurent Orseau, Richard Ngo, Miljan Martic)
- Specification gaming: the flip side of AI ingenuity (2020) (with Victoria Krakovna, Jonathan Uesato, Vladimir Mikulik, Matthew Rahtz, Tom Everitt, Ramana Kumar, Zac Kenton, Jan Leike)
Discussed in:
- Chapter 1: What Is AI?, A Brief History of AI