Imitation Learning
topic on 6 shows · 8 statements across 6 episodes
the Y Combinator Startup Podcast 
BG2 Pod 
Cheeky Pint 
American Optimist 
Latent Space 
No Priors
8 statements about Imitation Learning, every show
Dolgov: Autonomous vehicles require closed-loop RLFT, not just imitation learning
“And if all you did was kind of observing how other people drive when you trained the system, maybe observing, you know, just passively how people drive and how they interact maybe also, you know, driving the car yourself and then using imitative learning to tr…”
General Intuition's foundation agent runs purely on vision without reinforcement learning
“This is just a base model. There's no RL, no fine tuning. This model sees no game states. It is purely capable, not sequence acceptance. It's purely predicting the actions from the phrase. That's it.”
Making billions of gameplay clips playable bridges imitation learning to reinforcement learning
“Actually making every single clip on the platform playable at billions of clips scale is how we go from imitation learning to RL.”
Finn: Reinforcement Learning Outperforms Pure Imitation Learning in Robotics
“I think that reinforcement learning can play a very large role in it actually, in post-training. I think that online data from the robots which reinforcement learning allows you to use, Can allow robots to have a much higher success rate and also be faster tha…”
Wu: Generative AI has two distinct waves: imitation and reinforcement learning
“In, in generative AI, you know, everyone talks about kind of the whole generative AI wave. I actually think it's two waves. You know, the first wave is kind of what I'll call like imitation learning, and the second is, is more like RL, reinforcement learning.”
Vogt: Modern imitation learning solves unstructured home environments where classical mapping fails
“Using classical approaches, Where you have computer vision and you're trying to like reconstruct three D objects or fit to a map would make this like a really, really challenging and computationally intensive problem. Moving to more modern techniques like end …”
Gerstner: Tesla made more FSD progress in 12 months than past 7 years
“Tesla rebuilding their models, For how they do self-driving around imitation learning and all this interesting stuff going on over there. Like, you know, I think they probably made more progress in the last 12 months than in the last seven years.”
Gerstner: Autonomous driving will shift entirely to imitation learning over rules
“Well, here's where I think the world's going. We met with these robotics companies. We meet with Tesla, et cetera. Imitation learning, okay? They're not even gonna know what the stop sign is, or the traffic light is, or the dog in the street. They're not gonna…”