are there actual datasets to look at and info regarding how data was collected? all the sources on that page are just domain links but don't appear to point to the data making the claims?

4.7 accidents per million miles doesn't mean much if the cars are limited to specific roads or include test tracks that give them an advantage. the degree of variance in different environments would also need to be measured such as weather effects, road conditions and traffic patterns.

I'm all for autonomous driving, but its not like companies don't fudge numbers all the time for their benefit.