Cruse control requires you to pay close attention etc, but you can avoid making a lot of minor speed corrections, which makes long drives less taxing. Adaptive cruse control takes this one step further again reducing cognitive load. Lane following might not seem like much, but represents a huge number of tiny course corrections.
I could go on, but even if you can’t read a newspaper the reduction in stress and muscle fatigue really adds up.
Second, these systems are adding redundancy. As both the system and human driver needs to fail which means someone that’s only paying attention 98% of the time can still cover for 49 out of 50 times the system fails. Current data suggests a net gain in safety which is likely to improve as these systems mature.
Seems human nature to watch the car avoid 99 problems, and fall into a false sense of security and stop paying attention. Sleep, cell phone, talking to a passenger, or just day dreaming seems ever more likely the longer you are letting the car drive.
It's 2 people so far that lethally broadsided an 18 wheeler with a tesla on autopilot?
So an attentive human + FSD sounds great. Not sure humans are going to be able to stay attentive though. Sure at some point inattentive human + FSD will be safer than a stand alone human, but it doesn't seem like that's happened yet.
Telsa seems a bit evasive with the safety data. Maybe it's because human with autopilot off (but safety braking on) is just as safe as human with autopilot on.