Let's assume that after the full transition, 0 people will die from automotive accidents. This isn't the case, but it's close enough. Naively, that would save roughly around 30k lives a year: in other words, every single day self-driving cars are delayed kills a bit under 100 people. The dynamics of the transition don't matter too much, assuming everything is simply shifted forward a year: you just move one year from the totally untransitioned state to the totally transitioned state.
That's huge. Anyone who does something that delays the transition one day is basically pretty much committing the Pulse night club shooting and the Sandy Hook shooting.
Now, another question: if a self-driving Uber slams into a group of children crossing the street, how much do you anticipate that to delay the transition to self-driving cars?
There's also a hidden assumption that getting inferior products on the road as quickly as possible is a necessary step in order to start iterating as quickly as possible. A day of clean room testing is worth some small fraction of a day of real world testing. I don't quite dispute that point, but just stating it explicitly.