There are a few more categories of people who often do stupid shit, many of which are fairly mundane and far less easy to disregard (such as: momentarily distracted / semi-permanently fiddling with their phone / dangerously incompetent).
I don't think it's unfair to point out that humans, quite often, also suck quite badly at driving.
In general, I don't think human sucks at driving, but there are many times when any driver can be extremely dangerous for many reasons.
Many of those reasons are illegal (alcohol, fiddling with your phone) so it's not a question of fairness an automated program was supposed to be better than that.
I don't see anything unfair in requiring that automated driving which is supposed to not be afflicted from our human failing to not perform illegal and dangerous driving ...
My take is that both have spectacular edge cases / failure modes that are essentially laughable from the other's point of view. It's hard to compare the two, when all we're doing is cherry picking scenarios.
I could point to a double pedestrian crossing near my place where cars routinely speed. Various people have told me that the drivers cannot see the pedestrians, for a variety of reasons (buildings, cars, and especially a large tree near one corner), yet the human drivers are confident to go 10-20 kph over the speed limit through the area, despite this lack of information. I can give many similar examples - people really do suck at driving, they just tend to get away with it most of the time because rare events are, well, rare.
AI sucks for a variety of other reasons. The only real question is whether on average, over all scenarios and ratios of humans to AIs populating the roads, one sucks more than the other. It's something I don't think we have the answer to.
I don't think it's unfair to point out that humans, quite often, also suck quite badly at driving.