Curiosity

DeepCubeA, a deep reinforcement learning algorithm, can find the solution in a fraction of a second, without any specific domain knowledge or in-game coaching from humans.