Throws some good skepticism on the near-term timeframe estimates being thrown around for the advent of self-driving cars.
Indeed, but from the way Prof. Leonard explains these challenges, these are (primarily) problems that can be tackled by different teams at the same time. A bit like solving small puzzles that later make a bigger one. Then, most of the challenges seem to be on visual recognition... the cop says go on a red light, identify the cop, identify the hand signal, etc. I believe this will be sorted out relatively "fast". I think the biggest problem is human interaction, and the cross roads example is a great one. How it is eye contact with the other driver that determines what we do. In a world where 100% of traffic is not self-driving vehicles can be a bigger challenge.