by David Dudley, AARP
The car of the future arrives, a silver pod with the face of a robot puppy. It's not much bigger than a golf cart. The interior is a spacious but spartan affair: a thinly padded bench seat, a lone LCD screen where the instrument panel would be and, most strikingly, no steering wheel. Sitting down in what would be the driver's seat, I face a disconcerting void — an absence of controls.
The Google engineer beside me is careful to describe this as "a very early engineering prototype." Close inspection reveals it to be something less than a finished vehicle. The car's wide-eyed headlights and taillights are just painted on; the thin doors have no windows. It's like a prop car from a direct-to-cable science fiction movie.
But it's a very real and working test bed for the tech company's self-driving technology. Since 2009, Google has been testing a fleet of Toyota Prius and Lexus models. This unmanned prototype, the existence of which was revealed in May 2014, takes those cars to another level: a fully driverless vehicle. Since it lacks manual controls, it isn't legally permitted to venture out on public streets. So today we're taking it for a drive around a parking lot.
Or, rather, it's taking us. After we belt up, the engineer hits a button and off we go. The electric motor's turbinelike whine adds the proper sci-fi sound effects as we scoot around the lot. The vibe is akin to an amusement park attraction: You are, in every sense, merely along for the ride.
After perhaps 10 minutes, we are returned to our starting point. It's just a tantalizing taste of a future in which driving is something that is done for you, in a car designed to run entirely without a flesh-and-blood operator. A post-human automobile.
The implications of such a device have long excited safety advocates, who see a means of cutting the carnage on American highways. But the advent of self-driving technology will also affect the nation's growing number of older drivers. There are more than 45 million people in the U.S. age 65 or older, a figure that stands to grow by another 27 million by 2030. About 36 million current older drivers still hold valid licenses. About 80 percent of them live in car-dependent suburbs or rural areas, not cities with public transit. And nearly 90 percent say they intend to age in place.
What that adds up to is a looming legion of aging hands behind the wheel. For those whose independent living is closely tied to their ability to drive safely, self-driving tech is a future that can't come soon enough.
Bringing fantasy to life
It's a revolution long in coming. Sven Beiker, the executive director of the Center for Automotive Research at Stanford (CARS), observes that self-driving cars "have always been 20 years away." The automobile was barely out of adolescence when engineers began plotting ways to remove its most error-prone feature: the driver. At the 1939 World's Fair, General Motors' Futurama exhibit featured a domed cityscape plied by driverless jalopies. But bringing that fantasy to life has been a challenge. "We don't know when and how — and if at all — we will get there," says Beiker, whose research predicts how mobility patterns will change. "It's very hard to tell."
Driving is one of those deceptively complex human behaviors, like having a natural-sounding conversation, that machines struggle to simulate. The basics are easy — stay in your lane, stop at red lights, don't run into stuff. But maneuvering through traffic involves a dense interplay of anticipation, observation and decision-making. Is that guy turning left? Is she going to let me in? Is that a patch of ice, or water, or a pothole? Until recently, computers didn't have the processing horsepower.
The road to the autonomous car was going nowhere fast until the early 2000s, when the federal Defense Advanced Research Projects Agency (DARPA) offered $1 million to any team that could win a 142-mile off-road race for driverless vehicles. In the first DARPA Grand Challenge, in 2004, the most successful competitor managed only 7 miles. But a year later, five teams successfully finished. In 2007, DARPA added an Urban Challenge that required entrants to obey traffic laws and interact with other vehicles. Six teams pulled that off, and the dream of the driverless car was alive.
"That was a galvanizing event," says Chris Urmson, who was on the teams from Pittsburgh's Carnegie Mellon University that competed in 2004 and 2005 and won in 2007. "It felt like a robotics Woodstock."
A graduate student back in 2004, he's now a boyish 38 and the head of Google's self-driving-car program, his grad-student enthusiasm muted with a measure of Google corporate reserve: Sometimes, when I ask about certain details, he responds, amiably, "That's something we don't talk about." Urmson followed a well-worn path to Google after the DARPA races. Google's founders, Sergey Brin and Larry Page, declared that autonomous driving would be the company's first "moon shot" — an ambitious pie-in-the-sky project — and cherry-picked the top engineers in robotics, setting them up at a skunk works dubbed Google X.
Why does a search company start making robot cars? "Google is about solving big, important problems," says Urmson. "We have amazing research talent. We should be doing something with it to improve the world."
This sounds like familiar Silicon Valley–speak, but Google has a good case: More than 35,000 Americans were killed in car accidents in 2013, and 93 percent of traffic deaths can be at least partially attributed to human error. Urmson says Google is better positioned than a traditional automaker to crack the riddle of self-driving, because it's more about software than hardware: "When you look at what we're doing, on the surface you see a vehicle. But the heart of it is computer science."
Conservative but assertive
The people writing that code work in a building on the fringes of the Google campus in Mountain View, California. Outside sit several white Lexus SUVs, each crowned with a spinning silver appendage — a Velodyne laser turret, one of the array of sensors that allow self-driving cars to survey their surroundings with godlike omniscience.
Unlike the two-seater prototype I saw earlier, the Lexus fleet is street legal, as long as there's a human "safety driver" aboard. Today, Urmson takes the wheel, with engineer Shyan Izadian beside him with a laptop that monitors the car's systems. The interior is standard SUV, with a few exceptions: A screen on the instrument panel renders nearby cars in ghostly outline, and a fat red button sits on the console between the seats — a kill switch, in case of some unspecified robot-uprising situation. "We've never had to use it," says Izadian.
On Izadian's laptop, an elaborate digital map offers a closer look at how the car thinks. A yellow line snaking down the road indicates the intended route; other cars, bicycles and pedestrians are rendered as boxes of various colors. After a tone chimes, the Lexus surges out into the sea of distracted pedestrians and bicyclists that clog the streets of the Google campus and sets off on a 30-minute autonomous circuit of Mountain View. The steering wheel moves on its own, and the "driver" merely sits attentively, like a nervous parent. Izadian describes the car's driving style as "conservative but assertive." From the back, the only giveaway is the cheery female computer voice that alerts passengers of coming lane changes and crosswalks.
But there are other reminders that there's a cool and implacable alien intelligence behind the controls. When a light turns green, the Google car sits for a few seconds before proceeding. Urmson explains that the vast majority of T-bone collisions at intersections occur within the first moments, so the car is programmed to pause for two seconds. At another light, a gaggle of chatty bicylists sit in front of the car in the left-turn lane. When the light turns green, the bikers gab on obliviously. A human might hit the horn now, but the Google car just waits, with inhuman patience. "No need to create more road rage," says Urmson.
As it negotiates our human world, the car is also watching us: It's logging the behavior of other road users, so it can better predict how we'll act. "For example, we can learn what someone who's making a lane change looks like, and do a good job predicting what a normal lane-change situation looks like," Urmson says.
But sometimes there's no knowing what other people are thinking. As we accelerate over a freeway overpass, a beige Sentra pulls out in front of us. "Whoa," says Urmson, who takes the wheel and brakes sharply. A tone sounds as the auto-drive disengages. There's no screech of tires, but it's a close call.
"Well, that was interesting," Urmson says. Moments later, he reengages the auto-drive, and the chipper female voice announces the computer has resumed control. "That's good data," he says. "When something a little weird happens, we'll put a flag on it, and later we'll figure out what would have happened. The car really isn't ready to deal with that kind of reckless driving."
Google says that its cars have logged more than 700,000 miles without an at-fault incident. But that figure doesn't reflect how many times human safety drivers intervened, as Urmson just did. Self-driving skeptics also point out that the cars' driving turf is limited to areas, such as Mountain View, that have been digitally mapped to within an inch of their lives. A host of technical challenges still remain: Rain and snow can scatter the beam of the car's laser sensor, so the vehicle's bad-weather driving ability is an open question.
Urmson doesn't play down the amount of work remaining, but he sounds confident his team will meet Google's self-imposed deadline. "My boss said we need to do this in five years, and that was about a year ago," he says. "That's what we're pushing to. My clock is that my oldest son just turned 11. I've looked at teen-driver statistics, and they are not good."
How safe is safe enough? The stats for elderly drivers are troubling, too — after age 70, the rate of passenger-vehicle-driver fatal-crash involvements per 100 million miles traveled climbs sharply, according to the Insurance Institute for Highway Safety; by 80, drivers are dying at the same rate as rookie teenagers, though this is largely attributed to their greater vulnerability to injuries rather than an increased likelihood of getting into accidents. Ron Medford, a former deputy director of the National Highway Traffic Safety Administration, thinks the numbers would be worse if older drivers didn't "self-regulate" — take themselves off the road when their capabilities diminish: "People see it in their older parents. First, they don't drive at night. They don't drive more than 50 miles away, then 30 miles. Their lives just contract."
Google hired Medford two years ago as the program's safety director; he's also helping negotiate a daunting regulatory obstacle course. He's convinced that the lifesaving potential of the technology will be so dramatic that the industry will swiftly adopt it. "Never have I been as high on the possibility of a safety improvement in vehicles as I have with this," he says. "When I talk to seniors about this technology, they all say, 'Can I get it now?' "
Julie Lee, the national director of AARP Driver Safety, is a bit more measured in her optimism. The program she leads offers Smart Driver courses designed to refresh the skills of older drivers; recently, AARP Driver Safety collaborated with the MIT AgeLab to study how best to integrate new automotive technology. Loss of flexibility, vision and hearing all affect driving ability, but Lee stresses that such impairments can affect anyone, not just older people: "It isn't the age; it's the health of the driver." Lee was given a Google-car demo a few years ago. "It holds a lot of promise, but we have to be cautious," she says. "There are liability issues, licensing issues. We're a ways away from it."
How far? Bryant Walker Smith, a former transportation engineer who now teaches law at the University of South Carolina, thinks we could see autonomous vehicles in a decade, in part because of the demographic forces pushing the technology into the marketplace: "The demand is clearly there. Ten thousand people are turning 65 every day, and they're going to need something better."
Smith's research focuses on the murky legal issues surrounding the self-driving transition; as technology improves, he suspects we could see graduated licenses that permit people now considered unable to drive — because of poor vision or other impairments — to operate automated vehicles under limited conditions, such as within retirement communities or on special self-driving-only lanes. "The issue is, how safe is safe enough?" Smith says. "What accommodation are we willing to accept to give people access to mobility? It's a social question we'll struggle with."
Many automakers have been working on their own self-driving technologies for several years; others hurriedly jump-started programs in an effort to catch up. But car companies typically talk about self-driving as an assist for the driver, not a replacement. Google, on the other hand, has staked its claim on total autonomy. "They're a little more visionary than the established automakers," says CARS' Beiker. "They're looking at a different model. It's more a merger of personal mobility and public transportation."
One reason why, Urmson says, is that his team couldn't figure out how to "debug the human." Handing control of a speeding car back to a driver who's been blissfully zoning out for 60 miles is a challenge. There's also the risk of skill atrophy: A driver who rarely drives would eventually lose his or her fundamental competence. And a car that drives itself only most of the time, Urmson adds, "doesn't deliver on the mission of helping everyone get around."
The auto industry is leery of this idea, because cars that don't need drivers also won't need owners. Instead of private vehicles that sit unused 95 percent of the time, imagine fleets of shared autonomous shuttles that materialize when summoned, take you to your destination, then scoot off to assist the next passenger. Upending a century of private-car ownership would send shock waves through the economy, and society at large. It would also declaw one of the most-feared aspects of aging — the prospect of losing independence when we give up the keys.
The rest of my 30-minute drive proceeds without incident. We slow down in school zones, leave generous spaces when changing lanes, merge confidently. The Lexus chauffeurs me around Mountain View so serenely that I find my attention drifting. Where should I go for dinner? Boy, that's an old Volkswagen. How far are we from the beach?
This, says Urmson, is exactly what they're going for: "Our vision is that when technology works well, it kind of disappears. Look at the Google search page — it's a just a box that that you type into. Most users never contemplate the complexity of what happens behind that. I think that should be true of transportation. You express your desire, which is to get somewhere, and the technology just disappears."
You can see, sitting here, how tantalizingly close that vision is. And also how far away. Driving back to the Google campus, there's a truck partially blocking a parking lot entrance. We have plenty of room to drive around it. Something about the uncertainty of the situation, however, befuddles the car. It stops behind the truck and refuses to proceed, like a stubborn camel, until the engineer reluctantly turns off the device and takes the wheel again.