Episode
Leslie Kaelbling: Reinforcement Learning, Planning, and Robotics
1:01:33
Published: Tue Mar 12 2019
Description
Leslie Kaelbling is a roboticist and professor at MIT. She is recognized for her work in reinforcement learning, planning, robot navigation, and several other topics in AI. She won the IJCAI Computers and Thought Award and was the editor-in-chief of the prestigious Journal of Machine Learning Research. 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, Medium, or YouTube where you can watch the video versions of these conversations.
Chapters
In this AI podcast, Leslie Kaelbling, a professor at MIT and a roboticist, discusses her work in reinforcement learning, planning and robot navigation, and her contributions as an editor in chief to the prestigious journal, Machine Learning Research.
00:00 - 01:40 (01:40)
Summary
In this AI podcast, Leslie Kaelbling, a professor at MIT and a roboticist, discusses her work in reinforcement learning, planning and robot navigation, and her contributions as an editor in chief to the prestigious journal, Machine Learning Research.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
Stuart Russell discusses how philosophy contributes to artificial intelligence by addressing topics such as belief, knowledge and denotation.
01:40 - 07:49 (06:09)
Summary
Stuart Russell discusses how philosophy contributes to artificial intelligence by addressing topics such as belief, knowledge and denotation. He mentions the importance of knowledge expansion, reasoning with uncertainty, and the potential for creating robots that are behaviorally indistinguishable from humans.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The interviewee explores the history of AI, machine learning, and reinforcement learning from the 1950s up to now.
07:49 - 12:49 (04:59)
Summary
The interviewee explores the history of AI, machine learning, and reinforcement learning from the 1950s up to now. He discusses the roadblocks faced in the 80s and 90s with expert systems and symbolic computing.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The assumption that people can articulate how and why they make their decisions is flawed.
12:49 - 17:28 (04:39)
Summary
The assumption that people can articulate how and why they make their decisions is flawed. The experts' explanations for their actions are post hoc and not necessarily accurate.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
Markov Decision Processes (MDPs) model uncertainty in the future by allowing for various possible outcomes, but they also acknowledge that taking actions can change the state of the world, which is not deterministic.
17:28 - 21:36 (04:07)
Summary
Markov Decision Processes (MDPs) model uncertainty in the future by allowing for various possible outcomes, but they also acknowledge that taking actions can change the state of the world, which is not deterministic. MDPs account for uncertainty in the world, even if it is not fully known.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
In this podcast, the speaker discusses the concept of belief space and the problem of planning under uncertainty in real-world systems, where control is needed to manage beliefs in an uncertain environment.
21:36 - 31:44 (10:08)
Summary
In this podcast, the speaker discusses the concept of belief space and the problem of planning under uncertainty in real-world systems, where control is needed to manage beliefs in an uncertain environment.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The task of formulating human life as a planning problem is complex and cannot be solved by a single algorithmic solution.
31:44 - 38:18 (06:33)
Summary
The task of formulating human life as a planning problem is complex and cannot be solved by a single algorithmic solution. The problem becomes even more difficult when dealing with fuzzy goals and objectives.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The speaker believes that while we are far from understanding the full extent of perception and what it can do, the key to making progress is generating new ideas and implementing structures that are like the moral equivalent of convolution.
38:18 - 47:08 (08:49)
Summary
The speaker believes that while we are far from understanding the full extent of perception and what it can do, the key to making progress is generating new ideas and implementing structures that are like the moral equivalent of convolution.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The speaker discusses the importance of improving the peer review process in AI research and suggests organizing good public commentary to value the opinion of reviewers.
47:08 - 51:46 (04:38)
Summary
The speaker discusses the importance of improving the peer review process in AI research and suggests organizing good public commentary to value the opinion of reviewers.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The speaker emphasizes the importance of considering objective functions of value alignment when it comes to machine learning and artificial intelligence.
51:46 - 59:36 (07:49)
Summary
The speaker emphasizes the importance of considering objective functions of value alignment when it comes to machine learning and artificial intelligence. It is critical to have a longer research horizon to deal with the complexity of certain problems.
EpisodeLeslie Kaelbling: Reinforcement Learning, Planning, and Robotics
PodcastLex Fridman Podcast
The speaker asks the guest for their favorite robot from science fiction.
59:36 - 1:01:20 (01:44)
Summary
The speaker asks the guest for their favorite robot from science fiction. The guest responds by stating that they don't have one as their passion lies in engineering systems that work in the real world.