As a fan (of James Douma), I can’t help but buy into vision>radar. My Merc does phantom alerts all the time under bridges, tall items, unable to discern intelligently the difference between objects. Sure radar will give you better “vision” through rough weather but the goal here is to prepare the car for most situations most of the time, not edge cases all of the time, and more importantly to identify intelligently the object, velocity and so on. If object is identified as a human (erroneously) you want the car to err on side of caution. With time, and increased leviathan of data, you’ll see the Nets will learn to identify better.
That’s the bet He is willing to take. Bold but a bet nonetheless. Honda can equally bet on hydrogen, which in principle should work better, but unfeasible for wide scale adoption.
The solution should work pragmatically better —drastic reduction in cost that accompanies vision-only stance is a real driver of auto pilot/self driving adoption as a natural way of life. It has got to work. Much like a reusable rocket: it’s essential to wide scale adoption of jet-fueled travel both on and off the earth.