With 1.5 fatalities per 100,000,000 miles[0], the benchmark to meet is 99.9999985% of the time it doesn't kill you. Injuries are going to be a lot higher, obviously. Still, I think most self-driving enthusiasts underestimate the bar that needs to be crossed wrt safety. And general vehicle safety isn't going to remain stagnant. I think it's going to be a cost vs injuries tradeoff for quite a while until we get human-level or better self-driving safety in all circumstances.

[0]https://www.statista.com/statistics/193018/number-of-us-cras...