Google's car has been involved in at least 14 accidents in the last six years, all caused by humans. In 13 of the accidents, drivers of the other cars involved were at fault. In the remaining one, a human Google driver rear-ended another car while operating in manual mode.
Clearly, crashes caused by flawed humans is to be expected. Google's latest problem, though, is that its self-driving car isn't human enough. Engineers need to build more risk taking and aggression into the vehicle's code to keep it from getting stuck. In 2009, the car was paralyzed by humans inching up to a four-way stop, the New York Times reported. The Google car kept waiting for the drivers in motion to go, which left it sitting patiently at its own stop sign for far too long. Humans tend to inch forward or roll through stop signs to feel out what the other drivers are going to do. The Google car, on the other hand, is programed to stay still when anything else is moving.
Google's car also has a problem with negotiating near fixed gear bicycles, according to the Washington Post. Again, the four-way stop was the Google car's Waterloo. A bike rider met the car at an intersection while performing what's known as a track stand, without putting her feet on the ground. The motion of her bike rocking slightly back and forth was enough to keep the autonomous car glued to its spot, even though it had the right of way.
On the highway, where other drivers are more likely to be predictable, the Google self-driving car is doing great. The tech giant claims its vehicles have driven millions of miles accident free.