Humans average 60,000,000 miles between fatalitys [1]. It is a average of 1.25 million miles between injurys. 3 million miles is not a statistically significant amount of data to make a valid estimate, not even close.
That is not to say that safe testing can not proceed or that they are being unsafe in their testing or validation process, but people are bad with numbers, so it is important to realize the actual magnitude of the status quo and what we are comparing against.
Anecdotally, I feel much safer around AVs as a pedestrian or cyclist than human driven vehicles. Way more predictable, especially post pandemic when it seems like half of drivers have some behavioral issue.
That is not to say that safe testing can not proceed or that they are being unsafe in their testing or validation process, but people are bad with numbers, so it is important to realize the actual magnitude of the status quo and what we are comparing against.
[1] https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/... PDF Page 7