Lex Fridman Podcast

Pieter Abbeel: Deep Reinforcement Learning

Dec 16, 2018
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ANECDOTE

SpotMini Encounter

  • Pieter Abbeel met SpotMini at a Jeff Bezos event.
  • The robot impressively followed Bezos around.
INSIGHT

Robot Psychology

  • Despite knowing their mechanics, robots evoke a deep psychological connection. Even scripted robots like Pepper, designed with a childlike personality, can create this effect.
INSIGHT

Magic of RL

  • Reinforcement learning (RL) seems almost magical in its ability to learn from sparse, delayed rewards. It works by teasing apart which actions consistently lead to higher rewards and updates its policy accordingly.
Get the Snipd Podcast app to discover more snips from this episode
Get the app