Latent Actions in Robotics
The discussion delves into the challenges of applying latent actions learned from video games to real-world robotics. By leveraging expert-labeled data, they successfully mapped these latent actions to real actions, allowing an agent to learn and imitate behaviors from unseen videos. This innovative approach highlights the potential of using reinforcement learning in environments lacking labeled data.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Ashley Edwards - Genie Paper (DeepMind/Runway)
Related Questions