So, I'm sure you remember Alpha Go

That's the AI that beat a Go world champion.

Alpha GO was trained on a database of 100,000 Go games played by humans.

DeepMind has developed Alpha Go Zero…a new Go-playing AI that has no human training other than the rules of the game.

Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go. AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play. In doing so, it quickly surpassed human level of play and defeated the previously published champion-defeating version of AlphaGo by 100 games to 0.

AlphaGo Zero: Learning from scratch | DeepMind
We introduce AlphaGo Zero, the latest evolution of AlphaGo, the first computer program to defeat a world champion at the ancient Chinese game of Go. Zero is even more powerful and is arguably the strongest Go player in history. Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go. AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting…