I find the concept to be an AI viewed through the prism of human experience rather than an attempt to play the pure GAI. One of the theoretical advantage of fully self-aware AI is the controlled evolution. The AI does not need to mirror inferior humans, who were the result of a few billions years of natural selection, to evolve but by either insightful design or method of trial and errors achieve the result in a simulation in a much shorter span of time. Also, the motivation of "revenge" and the like look quite outlandish to a machine the motivation of which should in principle be incomprehensible to humans (and to nomads) due to being of different cognitive nature and level.
On a side note, I would suggest thinking of whether your AI is a network (machine swarm) with one master-mind or a group of semi- or fully autonomous independent entities.
