a generic version of their [Deepmind’s] algorithm, with no specific knowledge other than the rules of the game, could train itself for four hours at chess … and then beat the reigning computer champions – i.e. the strongest known players
The article references a recently published paper available on arXiv – Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm.
This appears to document DeepMind’s efforts generalise AI, allowing the same solution to be applied to different problems across multiple domains. The “fully generic” AlphaZero algorithm is applied to chess, shogi, and go “without any additional domain knowledge except the rules of the game, demonstrating that a general-purpose reinforcement learning algorithm can achieve, tabula rasa, superhuman performance across many challenging domains“.
*** UPDATE ***
AlphaGo Zero: Learning from scratch Deepmind blog post
Mastering the game of Go without human knowledge – Nature paper
The AI That Has Nothing to Learn From Humans
Deepmind announces AlphaGo Zero: Learning from scratch – OGS
South Korean pro strikes the first move on adapting Zero style