Chelsea Finn on Meta Learning & Model Based Reinforcement Learning
The Gradient: Perspectives on AI
00:00
How to Scale Up Data Collection
The idea of allowing people to tele-operate robots remotely potentially has a lot of promise for scaling up data collection. There's also been at Stanford as you probably know with Roboturk similarly scaling up demonstration data and yeah it'll be interesting to see if we can get to an amount of data that's useful and enough to learn this pre-trained thing that can then make it possible to not start from scratch in the future.
Transcript
Play full episode
