I think his main argument is that if it's measurably safer than a person, that's good enough. Dunno where the sweet spot is. My fear is that anything that isn

I think his main argument is that if it's measurably safer than a person, that's good enough. Dunno where the sweet spot is. My fear is that anything that isn't 100.00000% safe won't be allowed autonomy, it's easier to make a thing not happen than to make it happen. I expect that at this point it's all about data acquisition. Keep pushing out new releases as you learn something, hopefully the system needs less and less driver input. Crashes are probably pretty interesting to the engineers.