Source: đź“– 21 Lessons for the 21st Century
In 2017, Google DeepMind's AlphaZero beat the reigning computer chess champion, Stockfish. Out of 100 games, AlphaZero won 28 and drew 72, not losing a single game. AlphaZero was not taught anything by its human creators—not even standard openings. Instead, AlphaZero learned and trained by playing against itself using machine learning. Given that it didn't learn from any human players, many of its moves were considered unconventional—or you might say creative—if not downright genius.
The entire process of going from a state of absolute ignorance to acquiring a level of skill capable of making a mockery of the chess computer world champion took four hours of training.
Four hours.