Episode 25: Nicklas Hansen, UCSD, on long-horizon planning and why algorithms don't drive research progress
Generally Intelligent
00:00
How to Stabilize High Variance Q Targets With Data Augmentation?
There's something satisfying about seeing your machine learning algorithm working on a real robot. The only answer that we have so far is what you said, which is just do it a bunch of times and see what happens,. Yeah, it takes so many hours to do real robot experiments. That's true. Why did you decide to work more on robots more recently? I think specifically robots, there's something satisfying About Seeing Something Work In A Real World Is pretty satisfying.
Transcript
Play full episode
