Roads to a Universal World Model, Part 1: The Dreamer’s Road

The reinforcement learning path: learning by imagining

“The main idea of Dyna is the old, commonsense idea that planning is ‘trying things in your head.’” — Richard Sutton, SIGART Bulletin (1991)

What if a machine could practice inside its own imagination? Not in a hand-built simulation, where every rule is written by an engineer, but in a learned model of the world, one that the machine constructs from its …