cellery

I was passed two days ago in the bike lane by a self driving Uber - in a very complex, busy college campus environment. It went fine. The neat thing about deep-learning is that we're seeing a higher than linear - almost exponential rate of improvement with regard to complex scenarios that require drivers to take over.

There are many drivers that are incompetent and should probably not have drivers licenses, that safely pass cyclists and don't hit pedestrians. It is not realistic to expect bad drivers to somehow make exponential improvements - if anything they are getting worse. Then there are those drivers that do hit us... Because drivers can only get lucky for so long before their incompetence kills someone. I routinely see drivers veering into the bike lane before and after they pass me. People on their cellphones will occasionally drift into my lane. They will sometimes fly through a red light - like the time my girlfriend was t-boned by a distracted driver as she turned through an intersection. Heck, just the other day when we were driving home from another city on the interstate - about 90 mph, the car far in front of us continued driving straight, veering off a curve and nearly plummeted into a ravine. The driver corrected at the last second, tires screeching and all, saving themselves at the last moment. Let's not try to pretend that the bar of "safety" for self-driving cars is anything but pretty dang low.