Raw LLM Responses

I'm pretty sure that Tesla uses AI to guess the distance instead of using parallax to calculate it. There are 3 forward facing cameras, but they're very close together and at different focal lengths, so I'm not sure it can use that to effectively calculate parallax. Instead of switching back to radar+vision, I think they'd do much better to simply add another forward facing camera at couple feet to the side of the existing one with the same focal length. Farther apart allows for better depth perception at farther distances. A human would likely have only made the same mistake if they had one eye shut or blind and thus were relying on recognizing the object to then intuitively understand how far away it is based on its size. Computers can very accurately map the 3D world, but they either rely on techniques like LIDAR, or they rely on multiple perspectives of the same thing. Both can be very accurate. LIDAR could certainly be a good solution but I don't think the Radar+Vision was ever a good solution and Tesla was at least right to move away from that. Higher resolution cameras might also be needed. Honestly the camera resolution is kinda shit and things farther away are not that easily detected by it. Each of the front facing cameras are only 1.2 MP. I'm guessing that part of the reason for using lower resolution is it means faster processing by the AI, and maybe that's fine for lane detection and close up object categorization, but accurately guessing the distance to 3D objects seems kinda shit.