Jakob Foerster
TalkRL: The Reinforcement Learning Podcast
How to Influence the Learning Dynamics of Another Learning Agent
In the extreme case, you could imagine if I do this in a stock market environment, I might be able to use orders far away from the order book to influence the training process of other participants. It can also vastly accelerate that learning process and it can actually introduce a backdoor during the training of the victim agent. At test time remote control that agent to carry out a completely different policy.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
