The articles I've read on this have said it's murky because of obfuscation of reported data by self-driving companies.

The target article even notes that "while Waymo leads the industry in transparency, and robotaxi operators are required to report even the tiniest scrape to the National Highway Traffic Safety Administration (NHTSA), not every company voluntarily reports their total miles driven."

My sense is that there's a lot of massaging the data by autonomous car companies to make their data look good. The article and a linked article describes the comparisons as "human drivers in the same driving conditions" which is kind of vague to me and isn't the same as, say, taxi drivers to robotaxis. It's also robotaxis and not general driving conditions etc.

I always feel like these comparisons become murky in other ways. For example, if autonomous vehicles have lower accident rates in general, but fail in catastrophically bad ways when they do fail — ways that would never happen with a human driver (for example, by inexplicably driving into an avoidable semi trailer for no reason, killing the occupants) — is that better? What if you eliminate cases where the driver is intoxicated? Aren't those cases cases of malfeasance at some level, that could apply to autonomous vehicle use? Do you hold autonomous vehicle companies accountable for deaths, charging executives with manslaughter? How should we think about accountability versus safety?

I'm not trying to argue against autonomous vehicles. I just feel like these types of published discussions get oversimplified often in a way that seems to parallel similar discussions of AI in general, and seems sympathetic to their adoption rather than a rigorously skeptical appraisal.

They can also massage the numbers by only driving on the safest roads, then making a comparison based solely on miles driven.

The comparisons I've seen are with the general population, not professional drivers. How might that comparison look, since this is a paid service?