Tom McGrath describes how the neural network AlphaZero taught itself how to play chess without observing a human game.
See All 427 Episodes of "PNAS Science Sessions"