I doubt the ability of intelligence researchers to ultimately retain predictive capability of a self-correcting algorithm. If it is sufficiently powerful enough

I doubt the ability of intelligence researchers to ultimately retain predictive capability of a self-correcting algorithm. If it is sufficiently powerful enough to be considered able to learn, the complexity embodied in even simple biological creatures shows the inherent inscrutable nature of complex systems. I was stunned that Valiant seems to really believe a complex learning-capable system would continue to remain under human control. Though I am optimistic that we will continue increasing our capacity to control these types of systems (the Human Genome Project blew everyone away with their rapid advancements, so did Google's recent Go successes, etc), I am not so quick to jump to the conclusion that we will remain dominant of them in perpetuity.