Well, duh!
"allowing autonomous vehicles on the road when they're merely 10 per cent better than humans, versus waiting until “their safety performance is 75 or 90 per cent better than that of average human drivers”."
and " “under none of the conditions we explored does waiting for significant safety gains result in fewer fatalities.”"
So both the scenarios tested involved cars who drove better than humans and in no case would more fatalities be expected. Of course not! Proper test would require also a solution where autonomous vehicles are 10% less capable than humans!
And in any case how do you even gauge what is the performance of the "average human driver"??
Biting the hand that feeds IT
