Monday, March 04, 2019

AI: Alpha Go => AlphaGo Zero => AlphaZero (self-thought Chess & Go champion)

AlphaZero - Wikipedia

"AlphaZero (AZ) is a more generalized variant of the AlphaGo Zero (AGZ) algorithm, and is able to play shogi and chess as well as Go."

AlphaZero is a computer program or algorithm developed by the Alphabet-owned artificial intelligence research company DeepMind to master go, chess and shogi, by using an approach similar to AlphaGo Zero. On December 5, 2017 the DeepMind team released a preprint introducing AlphaZero, which, within 24 hours, achieved a superhuman level of play in these three games by defeating world-champion programs, Stockfish, elmo, and the 3-day version of AlphaGo Zero, in each case making use of custom tensor processing units(TPUs) that the Google programs were optimized to use.[1] AlphaZero was trained solely via "self-play" using 5,000 first-generation TPUs to generate the games and 64 second-generation TPUs to train the neural networks, all in parallel, with no access to opening books or endgame tables. After just four hours of training, DeepMind estimated AlphaZero was playing at a higher Elo rating than Stockfish 8; after 9 hours of training, the algorithm decisively defeated Stockfish 8 in a time-controlled 100-game tournament (28 wins, 0 losses, and 72 draws).[1][2][3] The trained algorithm played on a single machine with four TPUs

AlphaGo Zero - Wikipedia

AlphaGo - Wikipedia

AI versus AI: Self-Taught AlphaGo Zero Vanquishes Its Predecessor - Scientific American

mentioned in podcast:
Triangulation 387 Amy Webb: The Big Nine

This is not quite an AI except in name. Still an useful library

lhartikk/simple-chess-ai: A simple chess AI @GitHub

A step-by-step guide to building a simple chess AI –

No comments: