On the road with George Hotz’s $1,000 self-driving car kit

theverge.com

17 points·doczoidberg··3 comments

The big problem with ANY self-driving car is going to be the last .1%. Getting some sort of machine learning algorithm to drive well 99.9% of the time isn't that hard. Getting that same car to work well 99.999% of the time is where the challenge is.

The fact that Hotz seems like he doesn't recognize this fundamental problem is worrisome when thinking about the future of his company... and the field of self-driving cars. I wonder how quickly some sort of transit governing body is going to come down on him when they realize he can't possibly have done ANY kind of validation of the safety of his car.

It's a driver assist feature add-on after all. Human override is required at those .1% cases. Having said that, it's no different than when adaptive cruise control feature was first released or when GPS turn-by-turn navigation was first unveiled. They all warned you those are complementary driver assistances that don't replace driver to press the gas pedal or guiding directions because in rare cases(.01%?), those features don't work.

If we were to require a 100% accurate GPS navigation or a 100% working adaptive cruise control before they can be released to the market, we wouldn't have them even by now.

Five nines isn't enough, as evidenced by the articles lambasting Tesla for a single accident in many millions of aggregate miles. If you assume one decision per second--which is conservative--Tesla is already on the order of 10^9 decisions. Nine nines.

Reply on news.ycombinator.com