Episode 21: Chelsea Finn, Stanford, on the biggest bottlenecks in robotics and reinforcement learning
Generally Intelligent
00:00
Is There a Benchmark for Machine Learning?
I would not want to do reinforcement learning on the Shadow Hand. That sounds painful. I think that we should try to move towards more data driven and more automatic processes in can. There's an interesting field called unsupervised environment design. You need to build a simulator that's actually kind of close enough to the real world, which is hard in manipulation. But it's something that I think would be nice to get more people working with real data.
Transcript
Play full episode
