Also if we're going to need roads upgraded for autopilot it's going to be much much more limited. Roadworks take so long and we can barely keep roads together in some states.
There's more than one moving target here. One is when regulators allow the vehicles on the roads en masse. I think that should be at the point they're "better than the median driver." As far as user acceptance, I think once the price-point of automated driving as a service is cheaper than private ownership it will take care of itself. It will likely take a human generation or two to get the majority of people in autonomous cars.
My parents are quickly nearing the age that they will lose their independence/ability to drive. I would love the ability to put them in an autonomous car that is even a little bit worse than the median driver.
This is not the standard we need to meet. It's obviously the goal, but autonomous driving doesn't have to be perfect, it has to be better than the median human driver. Well, we need that plus a regulatory environment that handles the liability in such a way that results in countless lives saved, even though some might die due to imperfect autonomous driving.
I've long been of the belief that in order to get real autonomous driving we'll have to re-engineer major roadways. Today they're designed to maximize effectiveness of human sensors. I'm convinced that there must be a way to maximize effectiveness of non-human sensors without harming the effectiveness of human sensors. I don't know what that looks like. It might look like putting passive RFID in the reflectors on the road, or some other thing not yet imagined. But I do think that will ultimately be the future.