Alpha Zero and the future of artificial intelligence

Alpha Zero taught itself chess, without any human heuristics, without recorded games, or an opening book, by playing against itself

Years ago, I witnessed a thought experiment discussed by two old friends. One was an astrophysicist, the other a data scientist; neither was sober. The thought experiment: If you fed all relevant data into a computer, could the machine rediscover the laws of planetary motion by matching time periods, distances from the sun, velocities, et cetera, without human guidance?

I remembered that conversation when the self-taught Artificial Intelligence program, Alpha Zero, thrashed one of the world’s strongest chess engines, Stockfish 8. There were several amazing things about that match.

Alpha Zero won by an impressive score of 28-0 in a