Those self-driving cars
Innovation isn’t an easy road. Every improving technology occasionally lets us down, sometimes with deadly consequences. Look to Uber’s travail in Tempe, Ariz., as the latest example.
Maybe you’ve seen the video from an Uber self-driving SUV of that tragic moment: Elaine Herzberg, 49, is walking her bicycle across multiple lanes of roadway, moving from left to right. Cruising at about 40 mph in the far right lane, the Uber vehicle slams into Herzberg, killing her. A separate camera view shows the vehicle’s human safety driver, there to take control in an emergency, with eyes off the road and looking down before the crash. After impact, the safety driver looks up, stunned.
In assessing what went wrong, there’s a lot to unpack. The growing consensus is that the self-driving car’s technology, as designed, should have detected Herzberg in enough time to brake. That technology includes laser sensors, radar, and front, side and rear cameras—a combination engineered to see better than humans can, and react faster.
The collision happened at night, but a self-driving car’s robot eyes work best at night. Bad weather can interfere with the car’s vision, but the weather was clear. Then there’s the inattentive safety driver.
We offer condolences to the victim here, and to her family. But in this as in other avenues of progress, technology often fails before it can succeed. Failures inform. They tell engineers, scientists and innovators what needs to be reworked. And in the fledgling arena of autonomous car technology, on-theroad, real-world testing is the only way these cars can reach a point where they’re safer than having humans behind the wheel.
The human-driven auto, as its use expanded in the early 20th century, had a poor safety record that has markedly improved: In 1921, car use resulted in 24 deaths for every 100 million miles traveled, according to the National Highway Traffic Safety Administration. Today that measure is roughly one death for every 100 million miles.
It behooves Uber and other companies competing to perfect self-driving vehicles to learn from this death in Tempe. The New York Times reports that Uber has struggled to meet its goal of self-driving cars going a relatively scant 13 miles without the safety driver having to take control. Compare that to Waymo’s average of 5,600 miles before driver intervention, or GM-owned Cruise’s 1,200 miles before the driver has to intervene.
Driverless cars have remarkable potential to make streets and highways safer than they’ve ever been. They’ll take out of the equation the barroom drunk who insists on driving home, the reckless speeder who weaves in and out of lanes slalom-style, the distracted driver checking a text while merging onto the Kennedy. But those who would use the death in Tempe to slam the brakes on the development of these vehicles risk keeping all of us from that smarter, safer future.