This view about the inevitability of LIDAR seems to become increasingly anachronistic as time goes on. I have autopilot. Two nights ago it drove me from the entrance ramp to the exit ramp for my destination, approximately 30 miles, and I didn’t have to intervene once. The route involved three different highways, two overpasses, requisite lane changes, and even a brief stint through a traffic jam where it had to merge in heavy traffic. It was completely successful.
I fail to see how that is by any definition “behind”. I would be happy to replicate this and video it if you are skeptical.
Everyone agrees that you can do 99.99% of the driving without a LIDAR. That's exactly what you did yesterday in easy conditions (Highway driving is the easiest environment by far). The issue if you want to go to FULL self driving is that you need to go to 99.99999% of reliability which a lot of experts agree you will not be able to do without Lidar.
Here is the issue: Video cameras will work properly most of the time but once in a while the exact distance will be wrongly calculated as can be seen with all the Teslas crashing in trucks lately. A LIDAR gives you a way higher confidence on the distance of objects.
LIDAR gives you precision distance measurements. But LIDAR helps you little, if anything, with object classification. And object classification, the understanding what the car sees, is the big challenge with self driving.
Yes, with semi-intelligent systems it gives a head-start, as you can reasonably easy detect obstacles, but it does not help in understanding how to react to the detected obstacles.
So I would consider LIDAR rather a system which at the current state of development helps with collision avoidance rather than with true self-driving.
I fail to see how that is by any definition “behind”. I would be happy to replicate this and video it if you are skeptical.