The truth is 'ALLDAR", a combination of LIDAR and CAMERAS with an AI that selects the optimal solution as things change. Being tied to a single method give you problems when you hit the flaws inherent in that method alone. Use both or all three (if there is a number three??)
My take is that lidar is like ray tracing from decades ago. It was an expensive impractical "holy grail".
But presently lidar sensors cost as much as a car.
If the new lidar startups ship lidar sensors inexpensively at scale, then the economics will change and adding lidar to cameras and ultrasonics will help the cars drive.
So right now it would save more lives to put more people in reasonably priced cars with cameras.
I'm also reminded of the study where bicycle helmet laws kill more people than they save (because the benefits of having more people exercise outweighs the risks of unhelmeted riders)
Human beings are able to drive reasonably well (for some definition of “reasonably”), using only two visual sensors plus at least one audio sensor.
Elon seems to think that he can make a car that can do at least as well, with the same input limitations.
He may or may not be right, but personally I think it’s stupid to limit our devices to the same restrictions that we humans have.
If we can equip them to see outside of the human visual range of light perception, then I think we should probably do that. If we can equip them with more visual sensors, then I think we should probably do that. If we can equip them with sensors that have no human equivalent, then maybe we should do that, too.
I dunno. Maybe I’m just being stupid or something. I don’t get why he seems so violently opposed to LIDAR — other than the fact that if he added it now, then he would be proven to be a liar.