
New paper: Exploration potential
Measuring how much a reinforcement learning agent has explored of its environment class
New paper: Underprotecting statistical lives
Do we protect predictable lives more than statistical ones?
DeepMind collaboration
FHI and DeepMind are co-hosting monthly seminars aimed at deepening collaboration between their AI safety researchers
Jan Leike wins 'Best Student Paper Award' at UAI 2016
Leike presented "Thompson sampling is asymptotically optimal in general environments"
New paper: A formal solution to the grain of truth problem
Resolving a long-standing problem in game theory

New working paper: Openness in AI development
Prof. Nick Bostrom analyses how openness might impact on the development of artificial intelligence








