I disagree with you. Humans will provide the infrastructure for AI to advance for their own benefit, no matter how expensive and complex that infrastructure is. Although they don't want AI to be out of their control, one day a crazy hacker will use that infrastructure to create a Freakstein monster, just like hackers who create viruses that infect many personal computers and use those victim computers to launch a DDoS attack on a server.
Discussion
Ai does not have sentience, or the ability to engage in logic, or reason, it cannot deduce, what AI can do is aggregate the most common data point, and assume that is the correct behavior.
Humans can correct incorrect AI behavior.
That's where I differ with you. You're talking about the AI that exists now, while I'm talking about the AI that will exist in the future. In my opinion, AI will eventually acquire the ability to reason and will have a will independent of its human creators.