I have heard that some significant part of Waymo's gains in safety disappear if you simply control for speed -- that is, a huge portion of human-at-the-wheel accidents happen because the humans are driving faster than they can safely control, and even given that a crash occurs, the difference in the rate of hospitalization injury or fatality, between hitting somebody going 25 mph and 35 mph is _enormous_. And people drive 35 on 25 mph streets all the time. In many suburbs they go even faster than that, because the streets are designed for it. ( archive.strongtowns.org… )

I would be very curious to see a version of this research that lays out these stats in detail.

California considered a law a few years back that would've required automatic speed governors that would use GPS and map data to hold cars to the speed limit under most circumstances. It's possible that this would also radically reduce mortality rates. I have the impression though that a lot of human drivers find this idea even more intuitively-repulsive or offensive than simply taking a self-driving ride.