Learning from Mistakes
Sergey discusses the challenges of applying reinforcement learning in real-world scenarios, highlighting the limitations of current algorithms when faced with practical issues like breaking dishes. He emphasizes the importance of common sense in human learning processes and suggests that leveraging past experiences through multitask and meta-learning could enhance sample efficiency, allowing robots to learn more effectively without excessive trial and error.In this clip
From this podcast
Lex Fridman Podcast
Sergey Levine: Robotics and Machine Learning | Lex Fridman Podcast #108
Related Questions