Starcraft is a cool next step yet it kind of seems unfair in a way since the AI has distinct advantages in certain areas (namely, precision). A pro can play at

Starcraft is a cool next step yet it kind of seems unfair in a way since the AI has distinct advantages in certain areas (namely, precision). A pro can play at close to 300APM but are those actions always 100% precise when moving the camera/etc? It seems like unless you level the playing field in this regard it may be hard to take a lot of value from the AI winning since it may develop strategies that rely heavily on its advantages in those areas. Maybe that is completely fine though, or doesn't actually matter if strategic decisions trump precision. No Limit Poker is a pretty interesting problem, especially when it comes to reinforcement learning...can an AI learn to adapt to different play styles? Both of these games raise an interesting question as it relates to how exploitable a reinforcement learning AI will be in games where high risk decisions have major impacts and there isn't perfect information. Will the AI learn to have a more overall conservative approach to guard against these plays or will it itself exploit them.