AlphaGo Zero Learns Go from Scratch Without Human Data

DeepMind publishes AlphaGo Zero. The system learns Go from random play, no human data required, surpassing all prior versions in 40 days.