3 billion miles of two dimensional 1080P images is what they've got plus telemetry for speed, acceleration, etc. A lot of the imagery was acquired at night or in foul weather or taken by a camera with a dirty or broken lens.
If it turns out that lidar is necessary, there goes the value of the installed base of Tesla, vehicles. It would be like starting from scratch. Clinging to vision only AV is some variant of the sunk cost fallacy.
If you're thinking, surely they must've thought of all that and have a plan, I'd point to the design of starship and ask, is that ever going to be human rated?
They collect video, not images, along with other sensor and control data.
It’s not a sunk cost fallacy, it’s a technical strategy that is very logical and showing compelling results (though as of yet unproven for achieving robust L3 or L4 autonomy, demos notwithstanding).
If it turns out that lidar is necessary, there goes the value of the installed base of Tesla, vehicles. It would be like starting from scratch. Clinging to vision only AV is some variant of the sunk cost fallacy.
If you're thinking, surely they must've thought of all that and have a plan, I'd point to the design of starship and ask, is that ever going to be human rated?