That's a tricky one. Avoiding an accident for which you are not at fault could easily cause an accident for which you are at fault. Human drivers have a pretty bad record with that particular situation.

Also, there are some network effects at play here, the higher the percentage of self driving cars the more rare that situation should become.

I agree that there's a tricky calculus involved in some scenarios, but I think there's a much wider domain of not-at-fault scenarios where autonomous vehicles should be able to surpass humans unambiguously. Example: an autonomous car being tailgated by a human-driven car, with a stop sign or light up ahead. Different braking patterns have different likelihoods of causing an accident.

Autonomous cars really have to be able to handle these scenarios at least as good as, and ideally better than, humans: there's not a complicated split-second balancing act to worry about.

Tesla have stats saying their autopilot, for all its limitations, is already much safer than a human driver: https://electrek.co/2017/01/19/tesla-crash-rate-autopilot-nh...