*If you've got fifteen minutes to spare for the royal game of chess, you should watch this. It's Google's deep learner neural net intuitively smothering the most powerful chess engine in the world. Alpha Zero locks Stockfish into a position where every possible move it can make is either lousy or fatal. "Zugzwang."
*Commentators are describing the neural net's attack style as "strangling" or "smothering," and, although I'm plenty far from a chess expert, I can see that. Alpha Zero just doesn't number-crunch the situations on the board. It's doing what neurons do, it seems to resolves situations visually, like the nerves at the back of an eyeball. Watching these Stockfish-Alpha Zero games is like watching a powerful bulldozer utterly baffled by a drone with an overhead view.
*Stockfish plays great and never blunders, but it never has Alpha Zero even remotely in trouble. It's not even a sporting match; for a genuine fight, Alpha Zero needs to start the game, like, a knight down. After all, it mastered chess in a mere four hours of practice. There's no way the world's greatest chess-playing entity is breaking any sweat here. That "zugzwang," holy cow.
*After watching this, I'm wondering what Alpha Zero would do if you gave it a physics simulacrum of a human body and had it invent martial arts.
https://youtu.be/lFXJWPhDsSY
