Episode 22: Archit Sharma, Stanford, on unsupervised and autonomous reinforcement learning
Generally Intelligent
00:00
Adonimization - How to Make It Solve Things Faster?
Gail or S&C sometimes are not even able to solve the task. Usually your tasks can be solved in less than 1000 steps if you did it optimally. Since the agent doesn't know what's going on, it really just figure out where am I and what am I doing here? So it can take some more time. But generally ends up being like, I think it's trained to 60% times faster depending on the ability.
Transcript
Play full episode
