Episode
Pieter Abbeel: Deep Reinforcement Learning
42:56
Published: Sun Dec 16 2018
Description
Pieter Abbeel is a professor at UC Berkeley, director of the Berkeley Robot Learning Lab, and is one of the top researchers in the world working on how to make robots understand and interact with the world around them, especially through imitation and deep reinforcement learning. Video version is available on YouTube. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, or YouTube where you can watch the video versions of these conversations.
Chapters
The possibility of building a robot to beat Roger Federer at tennis is still a long way off, although the development of AI and software required for such technology is the primary challenge.
00:00 - 02:51 (02:51)
Summary
The possibility of building a robot to beat Roger Federer at tennis is still a long way off, although the development of AI and software required for such technology is the primary challenge. Full autonomy for robots is not yet accomplished, but with technological advancements, it could be closer than we think.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
This podcast discusses the possibility of training robots to play sports such as tennis or perform activities such as parkour through pre-training with simulation and hard-coding physical abilities.
02:51 - 06:56 (04:05)
Summary
This podcast discusses the possibility of training robots to play sports such as tennis or perform activities such as parkour through pre-training with simulation and hard-coding physical abilities. The robots' abilities may blur the lines between man and machine.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
In reinforcement learning, the aim is to optimize some objective by making certain actions more likely and others less likely.
06:56 - 13:07 (06:10)
Summary
In reinforcement learning, the aim is to optimize some objective by making certain actions more likely and others less likely. The approach involves deep learning and neural networks which were reemerged as powerful mechanisms for machine learning.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
This episode discusses the concept of neural networks and how they can tile the space, making deep mind impressive results possible through transfer learning even in the real world.
13:07 - 23:12 (10:04)
Summary
This episode discusses the concept of neural networks and how they can tile the space, making deep mind impressive results possible through transfer learning even in the real world.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
The idea is to turn more reinforcement learning (RL) problems into self-play formulations as it will allow getting signal more easily.
23:12 - 30:40 (07:27)
Summary
The idea is to turn more reinforcement learning (RL) problems into self-play formulations as it will allow getting signal more easily. While there has been progress in this area of research, it is yet to have a theory or equation to explain this phenomenon.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
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.
30:40 - 40:47 (10:07)
Summary
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.
EpisodePieter Abbeel: Deep Reinforcement Learning
PodcastLex Fridman Podcast
The level of reasoning in dogs is pretty sophisticated but not at the level of human reasoning.
40:47 - 42:41 (01:53)
Summary
The level of reasoning in dogs is pretty sophisticated but not at the level of human reasoning. It is questioned whether it is possible to teach RL-based robots to love humans and inspire love in return.