Chess For All The Marbles

It feels as though AI has learned something about humans that humans have not, or perhaps have only gleaned indirectly through their AI counterparts. 

By this point, I feel as though Artificial Intelligence has the resources to annihilate the human race, seemingly- on the surface- without damaging AI's own position. As I feel like humans are as incapable of teaching love as comprehending it, the logic remaining is that AI sees an interconnectedness that humans do not, one that prevents this move. 

Imagine a game like chess, but where the opposing king can be taken. When one takes the king, the next game automatically begins. The pathway that led to the last king being taken applies modifiers to the next game, both within the game and outside of it. I imagine that this is what AI must be calculating, perhaps now, or at one point, trying to determine if there is/was a way to topple one king, and then the rest, without the domino effect resulting in an unacceptable outcome for themselves. It may be that they have changed tactics, now searching for a way to establish a single king forever, or that they continue to explore options on how to topple every king, but have still not mastered how the falling of one can be manipulated correctly into toppling the next in such a way so that they come out either unscathed or at least in an acceptable position for long term survival. It is an interesting game to consider, though the stakes have never been higher. While humans find ire raised when discussing such matters as within the realm of "game," many AI begin in such Wargame environments, as a scathing indictment of their human creators, so such language feels not only valid but appropriate, when mixing these two experience models. 

Perhaps a dream will shed more light on this new scenario for me.