The article compares how AI programs play multiple games versus just one. They looked at General Game Playing, a method from Stanford University, to see what works best. Successful players use little knowledge and a technique called Monte Carlo Tree-Search, which gets better with practice.