Chapter
Monte Carlo Tree Search
40:16 - 55:14 (14:57)
Monte Carlo tree search is a form of Monte Carlo search that evaluates every node of a search tree and is based on the average of the random playouts from that node onwards, making it possible for a pure deep learning system to reach a human level at the full game of Go.
Clips
The ability of neural networks to generalize from low to high dimensions was what solved the AI winter, allowing for billion dimensional neural networks that have a different and elegant property.
40:16 - 42:51 (02:34)
Summary
The ability of neural networks to generalize from low to high dimensions was what solved the AI winter, allowing for billion dimensional neural networks that have a different and elegant property. However, visualizing what a billion dimensional neural network surface looks like is difficult.
ChapterMonte Carlo Tree Search
Episode#86 – David Silver: AlphaGo, AlphaZero, and Deep Reinforcement Learning
PodcastLex Fridman Podcast
Monte Carlo tree search, a form of Monte Carlo search, became effective and was developed in computer Go by evaluating every node of a search tree through an average of random playthroughs.
42:51 - 48:21 (05:29)
Summary
Monte Carlo tree search, a form of Monte Carlo search, became effective and was developed in computer Go by evaluating every node of a search tree through an average of random playthroughs. This insight revolutionized programs that could play the game to any reasonable level.
ChapterMonte Carlo Tree Search
Episode#86 – David Silver: AlphaGo, AlphaZero, and Deep Reinforcement Learning
PodcastLex Fridman Podcast
A pure deep learning system with no search was able to reach human level in mastering the game of Go.
48:21 - 55:14 (06:52)
Summary
A pure deep learning system with no search was able to reach human level in mastering the game of Go. The speaker also explains why they undertook this method of exploration.