One gripe here: everyone loves to use the fatality projections from RAND et al as the safety benchmark. Fatalities are the scariest thing but they’re an asinine metric — the lowest frequency of occurrence paired with the most (and most precarious) variables. AVs could improve on miles-per-fatality simply by calling police and EMTs faster than humans do after a collision. That’s a nice perk, but it’s not how we should test roadworthiness.

Waymo has already surpassed one of the projections in that same RAND report for lower-scale, more frequent driving errors. Certainly, we can and should have some discussions over why that does or doesn’t matter… and I think it would be a lot easier to have those discussions calmly and reasonably than to keep talking about fatalities.