For Tesla FSD to be truly autonomous, they will need to emulate this sort of performance from Waymo. In this report, Waymo reported 18 collisions per 6.1M autonomous miles driven. It took Waymo 21 months to generate this data, between 2019-2020. Data was collected with a human sitting in the driver’s seat to supervise the vehicle. By the end of 2023, Tesla will have approximately 2M vehicles operating with HW3/4 in the US alone. Assume each car drives on average 10 miles/day across the fleet, and assume that on any given day, 10% of the fleet is driving, Tesla could generate 6M autonomous miles in ~3 days. Tesla could get this data while the car still operates as a lvl 2 driver assistant system, as they’re doing today. This data could support a regulatory submission in less than a week. My point in all this: if Tesla can solve the technological challenge, everything else will happen FAST. Data will be collected FAST (and we won’t even know when Tesla is collecting data for the purpose of submission because they have the fleet-wide data and we don’t). And even if it takes the regulators time to review/approve Tesla’s submission, that is irrelevant because people won’t wait for the approval before buying FSD and Tesla won’t wait for the approval before pricing a fully-autonomous software appropriately. All of this can happen in a mere moments, and then, in an instant, Teslas everywhere would make human driving obsolete. And it won’t cost Tesla a penny in extra costs to get there. This is Tesla’s fundamental competitive advantage. THIS is the FSD Tesla bull case, in 1 post. arxiv.org/pdf/2011.00038