Jakob Foerster
TalkRL: The Reinforcement Learning Podcast
00:00
Open-Endliness: A New Approach to Multi-Agent Learning
There's a really sort of up-and-coming research area called unsupervised environment design, UD for short. And I've been fascinated by this question of how we can discover environment distributions that lead to specific translation results or allow agents to generalize the corner cases of the distribution. So being able to bridge the sim to real gap effectively is one of the key questions in bringing multi-agent learning systems to the real world.
Transcript
Play full episode
