AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
4min chapter
Jakob Foerster
TalkRL: The Reinforcement Learning Podcast
The Different Problems of Voltage Learning in Multi-Age Learning
In multi-age learning, one of the grand challenges is what policy do you play? It all depends on what others are doing. If you have a fully cooperative setting where you're able to control the policies so the way that other agents act for every single player in the team, then that effectively reduces to a single agent problem. In two-player zero sum, I can specify the requirement of having a Nash equilibrium. And if I find any Nash equilibrium, then I'm guaranteed not to be beaten by any training partner test time. That's why sort of competition and fully cooperative self-play are special cases of voltage learning.
00:00
Transcript
Episode notes
Get the Snipd
podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any
moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share
& Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode