Chapter
The Limitations and Possibilities of Autonomous Driving
30:40 - 40:47 (10:07)
The distinction between third person and first person is not important for autonomous driving, while effective simulation is crucial for self-play and imitation learning. It may be possible to teach RL-based robots policies of kindness, but it is difficult given the complexity of human policies that operate in the real world.
Clips
The development of autonomous systems such as self-driving cars would benefit significantly from effective simulation.
30:40 - 35:58 (05:17)
Summary
The development of autonomous systems such as self-driving cars would benefit significantly from effective simulation. An ensemble of simulators, rather than a single one, would be required to provide a sufficiently precise learning experience that can be transferred to the physical world.
ChapterThe Limitations and Possibilities of Autonomous Driving
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
The long-term optimization of human evolution has predisposed us to certain ethical behaviors, making it difficult for AI systems to naturally fall into similar policies.
35:59 - 39:16 (03:17)
Summary
The long-term optimization of human evolution has predisposed us to certain ethical behaviors, making it difficult for AI systems to naturally fall into similar policies. As AI technology improves, ensuring systems are indeed more capable and ethical than before becomes a complex and ongoing process.
ChapterThe Limitations and Possibilities of Autonomous Driving
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
Evolution built that into us to think about those things.
39:16 - 40:47 (01:31)
Summary
Evolution built that into us to think about those things. Steven Pinker talks about violence decreasing over time consistently. The long arc of history goes towards us getting along more and more.