Chapter
Clips
The history of breakthroughs in AI for game theory shows they typically occur in purely adversarial games, but the real challenge lies in understanding human gameplay and creating strategies based on that understanding.
1:30:10 - 1:31:48 (01:38)
Summary
The history of breakthroughs in AI for game theory shows they typically occur in purely adversarial games, but the real challenge lies in understanding human gameplay and creating strategies based on that understanding.
ChapterUnderstanding the Human Game
Episode#344 – Noam Brown: AI vs Humans in Poker and Games of Strategic Negotiation
PodcastLex Fridman Podcast
In order for machines to compete with humans, they must understand and adapt to the human way of playing the game, not just rely on their machine-like approach.
1:31:48 - 1:35:27 (03:39)
Summary
In order for machines to compete with humans, they must understand and adapt to the human way of playing the game, not just rely on their machine-like approach. While a chess-playing program may not need to mimic a human to beat one in a chess match, other games and interactions may require a deeper understanding of human behavior.
ChapterUnderstanding the Human Game
Episode#344 – Noam Brown: AI vs Humans in Poker and Games of Strategic Negotiation
PodcastLex Fridman Podcast
The true measurement of an AI's game performance is achieved by sticking it in games with varying skill levels to test whether it performs better than a strong or expert human player.
1:35:27 - 1:39:35 (04:07)
Summary
The true measurement of an AI's game performance is achieved by sticking it in games with varying skill levels to test whether it performs better than a strong or expert human player. This concept is similar to developing a self-driving car that should be tested on the road with non-expert drivers.