This evening in the car my six-year-old daughter, Alice, commented out of the blue that she wished our car could drive itself.
“I’d like that, too,” I said, and explained that scientists were, in fact, working on cars that could do exactly that, thinking of the Grand Challenges for driverless cars held by the Defense Advanced Research Projects Agency over the past few years.
“You should get one,” she said. I explained I couldn’t buy one yet, but maybe she could when she’s grown up.
“That would be cool,” she said.
Then I got home, started looking for a topic for this week’s science column, and the first item that popped up was John Tierney’s “Findings” column in the New York Times, headlined, “In the Future, Smart People Will Let Cars Take Control.”
Tierney quotes Sebastian Thrun, a computer scientist at Stanford University, who says, “Within five years, it’s totally feasible to build an autonomous car that will work reliably in several limited domains,” such tootling along a divided highway, inching along in heavy traffic, or finding its own parking spot in a mall or airport lot after dropping off the driver.
In other words, I may have been wildly pessimistic in telling Alice that she’d have to wait until she was grown up to have a self-driving car: I could indeed be driving…er, owning..one even before she earns her driver’s…er, owner’s?…license.
Currently, computer power doubles every couple of years. Consider chess: it used to be the big test of a computer’s “intelligence,” and matches between top players and computers like IBM’s “Big Blue” made the news. Then the computers started winning all the time. Odds are, as Tierney says, “human drivers will soon be outclassed by computers just as chess players are.”
You can see this progress in the results of each year’s DARPA Grand Challenge. In 2004, no car made it more than seven miles. In 2005, five cars finished a 132-mile course. And this year, six cars completed a 60-mile course in traffic, driving around an Air Force base in Southern California. Among other things, they found parking spots, obeyed stop signs, idled in traffic, yielded at intersections and merged into traffic at close to 50 kph.
“There was one accident and a few near misses, but the cars’ engineers are so bouyed by the results that they’re hoping the next competition will be a high-speed race on a Grand Prix course,” Tierney writes.
Of course, the biggest challenge preventing the birth of a new age of self-driving won’t be technicall: it will be the reluctance of humans to let go of the wheel. (And, indeed, if the idea of letting go of the wheel of a car at 100 kph in freeway traffic doesn’t alarm you, I hope I never have to pass you on the highway.)
As Tierney puts it, “There will be many, many car-computer jokes involving the word ‘crash.’”
But, he notes, cars don’t get distracted by cell phone calls and don’t drive drunk. And they don’t have to be infallible to be a lot better than good old human drivers, whose errors are responsible for more than 90 percent of traffic accidents.
With GPS systems, accurate internal maps and inertial sensors, an automated car will always know its position so precisely it won’t even need lane markings to navigate a road. And, of course, automated cars would communicate with each other, so they could run much more closely together more safely: no chance of a sudden stop by the one in front taking the one behind by surprise.
As Tierney notes, they wouldn’t even need traffic lights, because they could tell each other they were coming to an intersection, and would know who would get their first.
When will all this happens? According to Tierney, Dr. Thrun figures that within 20 years half of new cars sold will offer the driver of turning over some driving chores to a computer.
In 20 years I’ll be in the early stages of old geezerdom.
All you young whippersnappers will be grateful then that the car’s driving me instead of the other way around, I reckon.
And turn that dang stereo down!