A while ago I had one of those “Someone on the Internet is Wrong” arguments with the authors of an article arguing that we would need massively more evidence before we could conclude that autonomous cars are safer than those driven by humans. Rather than dig back to find those arguments again, I thought I’d <a href="http://I thought I'd link to this Bloomberg piece and, in particular the following passage”>link to this Bloomberg piece and, in particular the following passage
GM’s autonomous test cars were in 22 accidents in California last year, according to data from the state’s Department of Motor Vehicles … In a November interview, GM President Dan Ammann attributed the accidents to testing in a dense urban environment and noted the company’s cars weren’t at fault in any of the incidents.
Suppose that in any crash between autonomous cars and humans, each is equally likely to be at fault. What is the probability of seeing 22 crashes caused by humans and none by autonomous cars. Obviously, it’s the same as that of a fair coin showing 22 heads in a row, which is 2^-22 or about 1 in 10 million.
Of course, the drivers involved in the crashes aren’t likely to be a random sample of the population. As is standard in such things, the 80/20 rule applies: 20 per cent of drivers are responsible for 80 per cent of crashes and traffic infringements. THe 80/20 rule is derived from a Pareto distribution, and we can apply it a second time to say that 20 per cent of the remaining 80 per cent of drivers are responsible for 80 per cent of the remaining 20 per cent of crashes. That is, 36 per cent of drivers are responsible for 96 per cent of crashes. On that basis, it’s perfectly possible that the remaining 64 per cent of good drivers are as good as autonomous cars or even better.
It might also be argued that autonomous vehicles may fail in defensive driving, that is, in reducing harm in a crash caused by the failure of another driver.
Still, it seems pretty clear that autonomous cars are a lot better than the drivers responsible for most crashes and infringements. It isn’t that hard to identify a lot of these drivers before they kill themselves someone else, since prior driving record variables, particularly a driver’s prior traffic citation history, are the most consistent and powerful predictors of subsequent accident risk. Now that cars don’t need steering wheels or pedals any more, there’s no obvious reason to put people with bad driving records back in charge of them. Bad drivers should have their cars driven by robots.