I will only address the autonomy portion of the question.
The answer on who is killing that part is the industry themselves. They are doing that by choosing shadow driving AI as their method of gathering data, designing and testing these vehicles. The problem with that is AI has flaws. It missed corner cases and will take 1 trillion miles and over $300B per company/vehicle to get to L5. And most importantly, it puts shadow drivers and the public at risk needlessly. That because it has been proven that shadow drivers do not react well especially in critical situations and many, including professional, fall asleep at the wheel.
Why is this happening? Because the makers of these systems come from Commercial IT. Twitter, PayPal, AirBnB, Google etc. As such they have little domain or systems engineering best practice experience. And those who are relying on them, especially in government service, have even less experience in these areas. To make matters worse these folks come from a massive industry that is extremely successful in the types of things they make and generating massive revenue. That is causing a false sense of accomplishment and capabilities. The thought that they innovate everything and are now experienced enough to do anything. This does not mean they aren't gifted, talented, hard working or well meaning folks. But it does mean they don’t know what they don’t know and are over their heads. But neither they or their users or fans see that because of the massive echo chamber. Then comes the blessing and curse of AI. It hides the problems I just described to most people because it has learned enough for them to make it appear they are experts and are much closer to having autonomous than they really are. What is really happening is these situation are either gamed or these vehicles are driving well learned, mapped and benign routes. And by benign I mean there is no bad weather, difficult traffic patterns etc. However as I stated above that AI is far from perfect. Thus we have the perfect storm.
The answer is to use simulation. As MCity has stated it would drive that 1 trillion miles down by 99.9%. It would also drive down the time to L5 and avoid putting people at risk. Why isn't this happening? Because that same echo chamber doesn't think simulation can do what is needed. They are of course incorrect. Most of these folks have no idea what aerospace has been doing for the past 20 years. As such they do know know what simulation and simulators can do to mitigate the situation.
As an answer to this situation I am forming a group called Professionals for Safe and Responsible Autonomous Mobility. We are in the process of creating an international Simulation Association and holding a Simulation Trade Study. I have taken up enough space here. I would like to refer you to the two articles I have written below that explain all of this in detail and provides references for the data I cited.
Who will get to Autonomous Level 5 First and Why.
Stop relying on AI to make Autonomous Vehicles — You are wasting time and risking lives