AlphaZero is given the rules of chess and 4 hours playing itself in order to become the top-rated chess program in the world!?
Impressive!
Is it just my impression, or is DeepMind pretty much the only commercial outfit publishing innovative AI stuff?
I can't wait to see them tackle textual / natural language tasks.
P.S. Ruy Lopez (e4 .. Nf3) is the best opening for white, here's the paper: https://arxiv.org/pdf/1712.01815.pdf