Pete Florence: Dense Visual Representations, NeRFs, and LLMs for Robotics
The Gradient: Perspectives on AI
00:00
Is Language Models Really Helping With Vision and Action Fusion?
There's some very interesting ideas. I'll say one is codus policies, you could imagine it may be being extended to also kind of do the deeper things on the perception level side. And then there's also this just general idea that there's this really nice paper that I'm thinking about from a colleague of mine, Igor Mardash. So even though language models don't inherently have too much of this low level touching the world and seeing the world and having the world respond to you information inside of them,. they could somehow be used through additional training, additional data to then help with these kinds of you know, dexterous, for example, manipulation problems.
Transcript
Play full episode
