Games are traditionally used as markers of progress in artificial intelligence. Most of the previous approaches focused on a single game until AlphaZero mastered three different games. However, these were perfect information games, and the extension to imperfect information games, like poker, is unclear.
Technology Org logo
Google Play icon
Player of Games: Improving Guided Search, Learning, and Theoretic Reasoning
Share
Posted Yesterday
This news or article is intended for readers with certain scientific or professional knowledge in the field.
Games are traditionally used as markers of progress in artificial intelligence. Most of the previous approaches focused on a single game until AlphaZero mastered three different games. However, these were perfect information games, and the extension to imperfect information games, like poker, is unclear.
Games are traditionally used as markers of progress in artificial intelligence.
Image credit: geralt via Pixabay (Free Pixabay licence)
A recent paper by DeepMind introduces Player of Games, a new algorithm that generalizes the class of games in which strong performance can be achieved.
It uses self-play learning, search, and game-theoretic reasoning. Player of Games is the first algorithm to achieve strong performance in domains with both perfect and imperfect information. It uses using a single algorithm with minimal domain-specific knowledge to master fundamentally different games: chess, Go, poker, and Scotland Yard. The proposed approach is an important step towards general algorithms that can learn in arbitrary environments.
Games have a long history of serving as a benchmark for progress in artificial intelligence. Recently, approaches using search and learning have shown strong performance across a set of perfect information games, and approaches using game-theoretic reasoning and learning have shown strong performance for specific imperfect information poker variants. We introduce Player of Games, a general-purpose algorithm that unifies previous approaches, combining guided search, self-play learning, and game-theoretic reasoning. Player of Games is the first algorithm to achieve strong empirical performance in large perfect and imperfect information games — an important step towards truly general algorithms for arbitrary environments. We prove that Player of Games is sound, converging to perfect play as available computation time and approximation capacity increases. Player of Games reaches strong performance in chess and Go, beats the strongest openly available agent in heads-up no-limit Texas hold’em poker (Slumbot), and defeats the state-of-the-art agent in Scotland Yard, an imperfect information game that illustrates the value of guided search, learning, and game-theoretic reasoning.
Games have a long history of serving as a benchmark for progress in artificial intelligence. Recently, approaches using search and learning have shown strong performance across a set of perfect information games, and approaches using game-theoretic reasoning and learning have shown strong performance for specific imperfect information poker variants. We introduce Player of Games, a general-purpose algorithm that unifies previous approaches, combining guided search, self-play learning, and game-theoretic reasoning. Player of Games is the first algorithm to achieve strong empirical performance in large perfect and imperfect information games -- an important step towards truly general algorithms for arbitrary environments. We prove that Player of Games is sound, converging to perfect play as available computation time and approximation capacity increases. Player of Games reaches strong performance in chess and Go, beats the strongest openly available agent in heads-up no-limit Texas hold'em poker (Slumbot), and defeats the state-of-the-art agent in Scotland Yard, an imperfect information game that illustrates the value of guided search, learning, and game-theoretic reasoning.
You must be logged in to post a comment.