I've read all that stuff before and none of it addresses the fundamental problems at issue here. We are realistically still decades of intense R&D work away from having an autonomous AI that can respond with enough flexibility to rapidly changing real world situations.
We are realistically closer to autonomous AI than you imagine. Pessimism, budget planning that didn't anticipate it, ethical concerns and human employment are slowing down the development.
If you are not convinced, you can look for "waymo simulation" in your favorite search engines.