AI agents trained in simulations that differ from the environments where they are deployed sometimes perform better than agents trained and deployed in the same environment, research shows. Read More in  Artificial Intelligence News — ScienceDaily