That was the first thing that occurred to me, too. As the earlier discussion about chess suggests, interactions with other intelligent agents complicate matter

That was the first thing that occurred to me, too. As the earlier discussion about chess suggests, interactions with other intelligent agents complicate matters. An intelligent agent shouldn't just attempt to maximise the number of possible states -- such an agent may well choose to be passive, and allow the other agent to run the show, which isn't a very useful outcome. Rather, an intelligent agent should attempt to maximise the number of outcomes WHICH IT CAN DETERMINE. And that is the will to power.