Lex Fridman had Jim Keller on his podcast and they talked about his time at Tesla and his work on auto pilot and autonomous driving in general. Keller was so ca

Lex Fridman had Jim Keller on his podcast and they talked about his time at Tesla and his work on auto pilot and autonomous driving in general. Keller was so cavalier about the solvability of self driving. I was wondering how much of Elon’s boundless optimism and reality distortion field had rubbed off on him. His basic point was that with enough high quality data the problem can be solved, but never really addressed the algorithmic complexity of driving. I used to be very bullish about it myself, but after paying more attention to my own driving and how many edge cases and corner cases there are, I am much less convinced.