I mean, humans have only their eyes. And most of them intentionally distract themselves while driving by listening to music, podcasts, playing with their phones, or eating.
But (some) humans have the ability to handle difficult situations, and no autonomous system gets anywhere close to that. So this is more of a "robots handle the easy 80% better, but fail hard on the rest of the 20%". Humans have a possibly worse 80% performance, but shine in the 20%.
If you include minor fender-benders and unreported incidents, estimates drop to around 100,000–200,000 miles between any collision event.
This is cataclysmically bad for a designed system, which is why targets are super-human, not human.