“The original motivation for the Genie project, started in 2022, was to solve the problem in reinforcement learning of needing unlimited, diverse environments for training agents.”