He not wrong though. Most people made bets that self-driving with basic ML can be done with LIDAR by 2020 a year early is fine. And he's right about LIDAR. Theres a bunch of paper outlining that pseudo LIDAR can make an equal projection about depth based on the KIIT metric LIDAR measures themselves against. That with a fraction of Tesla's data. Its overzealous but it if he shows it working 99% its going to cause a lot of heads rolling.
He also the annoyed the hell out of ULA, Boeing, Roscosmos, and ESA.
The next few years will definitely be interesting for Tesla and the self-driving industry as a whole.