The difference is in how you measure "better than humans." Which also means looking closely at how we measure how "good" a human driver is.

It's the same argument as "what is AI" which inevitably devolves to "must be demonstrably smarter than humans in every way before we consider it to actually be intelligent."

I also suspect that the average human driver is getting worse not better. We have way more distractions available in the cabin than in the past mostly due to cell phones. I also notice a lot of younger people are either forgoing driving altogether or getting their licenses much later due to the availability of uber/lift as an alternative so would imagine they have less experience.