By now, it has become a go-to quip for Chris Urmson, father of a 12-year-old son and, in effect, father of Google’s self-driving car.
His design team’s goal, he tells audiences (and told me in an interview Thursday at the Austin Club), is to make sure his boy does not need a driver’s license when he turns 16. In other words, for autonomous cars to become an on-the-street fact of American life by sometime in 2019.
They are of course already something of a fact of life in Austin, where, as I’ve reported before, the cars have been in testing mode since last year. But he means on the street, commercially.
It’s a good line, simultaneously playing to most parents’ wariness about releasing their precious babies to the mean streets behind the wheel of a couple of tons of internal combustion, while also confronting widespread skepticism about the prospects for self-driving cars.
Urmson, the California-based director of Google’s self-driving car program, was in town for South by Southwest Interactive, but squeezed in a half-hour or so to talk with me.
He has heard all the arguments from the skeptics, including reporters: No one will be able to afford a self-driving car, given all those lasers, cameras and semiconductors. You’re just putting dozens of new devices on a car that will eventually malfunction and confound auto mechanics. People love to drive, to be in control rather than be passive occupants of a car. And self-driving cars will never take over the road because there will always be holdouts for the old way.
But Urmson said the point isn’t necessarily to have 100 percent of cars autonomous, or any particular percentage. Whatever share ends up being driverless, he and others argue, that’s a slice of people who will be less likely to cause a wreck.
“It needs to work in the world as it is now,” he said Thursday. “Otherwise, you end up in some sort of chicken-and-egg place where it’s ‘we’re not going to let them out until they’re all self-driving.’ And then it never happens.”
So Google (along with the numerous other car and high-tech companies that have been working on driverless vehicles) set out to make its vehicle the ultimate defensive driver. I’ve written before about how poky the cars can be, how cautious the “decisions” made by the computer metaphorically behind the wheel. And it’s working so far.
In 1.4 million miles of testing here, in Google’s Silicon Valley home base of Mountain View, Calif., and, most recently, Kirkland, Wash., near Seattle (“it rains more there,” Urmson said, providing a needed experimental condition), the cars have been involved in 18 accidents. Sixteen were the fault of the other driver, and one was caused by a Google human driver who had the car in manual mode. The sole mishap laid at the feet of the computer occurred this Valentine’s Day, a paint scraper in which the Google car anticipated that a bus attempting to pass it on the left would stop.
Urmson said the crash occurred at about 2 mph. There were no casualties. So, one crash in 1.4 million miles. At a typical 12,000 miles a year of driving for a person, that would be about 117 years of accident-free driving, interspersed with one fender bender.
As for the skeptics’ bill of particulars, Urmson has answers. The cost of the vehicles, he said, will come down into a marketable range when mass production starts. And the equipment doesn’t have to be installed only in a high-dollar vehicle like the Lexus SUVs Google is using for testing, he said, but could be fitted onto much cheaper models.
Yes, he acknowledged, “there’s certainly more stuff to break” on a self-driving car, stuff that would be complicated.
“But one thing the car companies seem to be really quite good at is building interchangeable modules,” Urmson said. “I can imagine that carrying forward as the car goes from a prototype to something that’s out in the world. And it would need to.”
As for the thrill of the road, Urmson noted the well-camouflaged pleasures of Interstate 35 just a few blocks east of where we were talking.
“That’s not a joy,” he said.
Then there are the mobility advantages a self-driving car could offer to people who can’t drive a car because of disabilities or old age.
“Having those people be able to have the same freedoms that you and I take for granted seems really important,” Urmson said.
And there’s safety. Last year, about 38,000 people died in America in vehicle accidents and, Urmson said, human error was the cause in 94 percent of them. Preventing some of those deaths is probably worthy of Google’s time and energy, and, later, the public’s money. Urmson’s son might be one of the beneficiaries, although I can envision perhaps some disagreement around a Bay Area dinner table when the subject of getting a license arises.
“Geez, in 3½ years he’ll be 16,” Urmson said last week. “I gotta stop saying that.”