I think the argument, equipment-wise, is that humans drive ‘safely’ with only two cameras on a swivel.
From a sensor standpoint, advantage already goes to the car/computer on that one.
Obviously we have very good pattern-detection and image recognition software running upstairs. Right now, advantage: human.
Our reliability/attention, not so good. Advantage to the computer on that one.
Elon’s argument is that the second point is just a software/computational problem, and once it’s solved, will result in a superior driver than humans.
Personally, I think he underestimates the timeline that will happen in, and the amount of computational power required.