As HAL 9000 used to say in the movie 2001: A Space Odyssey, "it can only be attributed to human error." Nowadays, we have real A.I., and the people should still be blamed when A.I. freaks out and starts killing people.
Police in Tempe, Arizona, said the "safety driver" behind the wheel of that self-driving Uber that killed a pedestrian in May was watching TV on her phone at the time of the crash. (Specifically, she was watching "The Voice" on Hulu.)
The department's 318-page report claimed that driver Rafaela Vasquez kept taking her eyes off the road before hitting and killing Elaine Herzberg. The crash was avoidable.
Their conclusions were based in part on in-car video, which showed Vasquez glancing up occasionally, but mostly looking down at her phone. The car was in self-driving mode at the time of the crash.
After the accident, Vasquez claims she was not using her phone in any way. But Hulu's records showed that she was streaming "The Voice" at the time of the crash.
Police estimated that in order to avoid hitting Herzberg, Vasquez would have had to brake for 5.7 seconds. But she was watching TV for 5.2 seconds of that time, looking up only at the last half-second.
The cops' conclusions match NTSF's -- Vasquez was to blame.
The crash, which represents the world's first fatality from a self-driving car, highlights an underappreciated challenge with self-driving cars: human nature.
Specifically, the better they get at driving, the more likely humans are to assume they'll handle everything and not pay attention.
This is similar to the psychology behind other safety features like seat belts. The increased survival rate from drivers wearing seatbelts are partially offset by the fact that people tend to drive faster because they feel safer wearing the seatbelts.
This phenomenon is described by an idea called the risk compensation theory. The safer people feel, the less careful they are.
Someday, no doubt, self-driving cars will be amazingly safe. But in the transition, we do need to hold safety drivers totally accountable, and to constantly remind them that they're every bit as responsible as any other driver.
And we need to figure out some way for safety drivers to remain vigilant, even after thousands of hours of safe autonomous driving.