Will regulators allow machines to kill people? Human drivers killing themselves and others are a part of human nature, but will politicians and the public accept that people get killed through no fault of their own by mistakes machines make?

Even ignoring the answer to that question is yes for other industries. They are allowing SDVs even after the Uber/Phoenix incident, so we already know the answer to this question.

> Human drivers killing themselves and others are a part of human nature, but will politicians and the public accept that people get killed through no fault of their own by mistakes machines make?

This has already happened with manufacturer defects. It's just more of the same. As long as a clear liability boundary can be established in principle, there's no problem.

If autonomous driving can be delivered with accident-causing failure rates similar to those of general manufacturer defects (i.e., much lower odds than accidents caused by human error), then I think that would be sufficient. But if say we go from ~35k human-caused road deaths/year to ~10k autonomous-driving caused deaths/year, that may be seen as unacceptably high.

> But if say we go from ~35k human-caused road deaths/year to ~10k autonomous-driving caused deaths/year, that may be seen as unacceptably high.

Sure, but there wouldn't be an outright ban. They'd instead establish higher standards that need to be met.