IPAB Seminar - 5th May 2022

Title: Towards embodied intelligence

Abstract: Enabling simulated and real-world embodied agents to think and move like animals and humans is one of the shared goals of AI researchers, roboticists and the computer graphics community. In my talk I will review some work that we have done towards this goal. In the first part I will discuss work towards building intelligent simulated humanoid characters that possess locomotion and manipulation skills, that can see and remember, and that interact with each other. I will bring together results from several studies, and explain how large scale RL, imitation learning, hierarchical skill representations, and multi-agent training algorithms can work together to this end. In the second part I will discuss how some of these results can be brought to bear on real robots. And, time permitting, in the third part, I will discuss some of the additional challenges associated with robot learning in the real world, and review some algorithmic progress that has been made over the last few years, which makes me optimistic that this goal may eventually be achievable.