Learning to Play, Imitate and Collaborate with Pesky Humans

Speaker: Dr Sam Devlin, Microsoft Research

Sam Devlin shares design patterns he has learnt researching and developing machine learning for non-player character control from collaborations on Age of Empires 4, Bleeding Edge, Minecraft, and more. The talk cuts through the research jargon and AI hype to a core set of three fundamental approaches: learning to play by reinforcement, imitating by behaviour cloning, and combining these to enable collaboration with any human player.

Date:
Haut-parleurs:
Sam Devlin
Affiliation:
Microsoft Research

Regardez suivant