Lex Fridman Podcast #144 – Michael Littman: Reinforcement Learning and the Future of AI
7 snips
Dec 13, 2020 In a captivating conversation, Michael Littman, a computer scientist from Brown University, shares his insights into reinforcement learning and its transformative potential in AI. He delves into the interplay between sci-fi and robotics, touching on humorous robot-human interactions. Michael also discusses the challenges of AI's existential risks, the creative process behind parody songwriting, and the evolution of AI through self-play. His reflections blend technical depth with light-hearted anecdotes, creating a thought-provoking dialogue about the future of technology.
AI Snips
Chapters
Books
Transcript
Episode notes
First Computer and College Choice
- Michael Littman got his first computer, a TRS-80 Model 1, at age 13.
- He chose Yale for college specifically because it offered a computer science major.
Tesaro's Neural Net Mastery
- Michael Littman considers Jerry Tesaro a "neural net whisperer".
- Tesaro has an ability to coax neural networks into performing feats others find impossible.
Value of Worst-Case Analysis
- Worst-case analysis in algorithms provides valuable modularity for system design.
- It allows for predictable integration, unlike relying on performance in specific practice.
