AI Trends 2023: Reinforcement Learning - RLHF, Robotic Pre-Training, and Offline RL with Sergey Levine - #612
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Is There a Parallel to the Learning From Videos in Robotics?
The learning from videos area is something where there's been a lot of action. Eagle 4D data set has first person videos of object manipulation from humans, from humans wearing cameras. In parallel to some of this work out of meta and some of these other places that focus on videos, in my group we've been focusing on using robot data sets. We're not actually sure whether those visual features encode something about the behavior itself or merely something about perception of objects.
Play episode from 30:30
Transcript
