AlphaZero Is Back!

Ding, ding, ding! As I made my way through this morning's reading list on chess topics, the bells were sounding everywhere. Almost a year to the day after their first shock announcement, Google's Deepmind had just released more news about AlphaZero, by all reports the strongest chess player ever.

The news was propagated via Science magazine -- see Table of Contents : December 07, 2018 (sciencemag.org; cover shown on the left) -- which included three articles by world class authorities on computer chess:-

• Editorial: 'Chess, a Drosophila of reasoning' by Garry Kasparov
  
  
• Perspective > Computer Science: 'Mastering board games' by Murray Campbell
  
  
• Report: 'A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play' by David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, Timothy Lillicrap, Karen Simonyan, Demis Hassabis

The description for the cover of Science said,

Starting from random play and given no domain knowledge except the game rules, the AlphaZero program taught itself to play chess, shogi, and Go, defeating a world champion program in each game. Blue translucent pieces represent AlphaZero's possible moves; percentages indicate the predicted outcome. A single algorithm that can master several complex problems is an important step toward creating a general-purpose machine learning system to tackle real-world problems. • Image: DeepMind Technologies Limited

This is too much new material to digest in the time available for a simple blog post, so I'll come back to the subject as soon as I can.