We are building learning algorithms that allow agents to achieve goals in big worlds.