The headline is misleading (click-bait). Reading the headline, I imagined they have simulated the entire economy of the world and have used reinforcement learning to find optimal policies, but they only simulated 4 workers and are making generalization to the entire economy.