The first automated driving solutions are, and will continue to be, used in situations where humans are at their best - the easiest driving conditions. So it's misguided to compare their performance to humans in non-perfect conditions. If an AI driver is driving in other than perfect conditions, and there is a crash, apologists can and will say "well, the idiot human should have switched off the AI". The AI is inherently considered better because of mental gymnastics that deny responsibility. This is an example of how humans are exploiting defects in other humans' reasoning, not how AI is superior.