Thrun met Google cofounder Larry Page through the DARPA challenges, and after the race Page approached him about coming to work for Google on some groundbreaking projects (including Street View). Thrun accepted. “Google is run by exceptional, visionary leaders with very deep technical curiosity and an ambition and a calling that goes way beyond organizing the world’s information,” Thrun told me of his decision. “It became clear that with the work we did at Stanford University, in my lab…there was an opportunity to fundamentally transform society.”
It made sense to go to Google, he said, because the challenge of self-driving cars is “not an automotive problem, it’s not an engineering problem in the classical physical, electronic, or mechanical sense. It’s really a problem in terms of computer science, and Google is really one of the leaders in the world on that, with a scale of computing and staffing that was hard to get at any university.”
Thrun assembled a team of twelve engineers, including some of the people who worked with him on the DARPA challenges.10 Soon, they were loading up Toyota Priuses with LIDAR sensors and trunks full of computers. They began testing the cars on public streets, with human drivers behind the wheel in case they needed to take control of the car in a dangerous situation. Early on, human intervention was common. “In the beginning,” Thrun said, “we were able to drive maybe in the order of ten miles between critical incidents.” The more they tested the cars, the more unexpected situations they encountered and programmed the cars to handle properly—when people ran out into the street, for instance, or when a driver in front of them suddenly changed lanes.
In fewer than three years, the team had advanced and tested the technology far more thoroughly than anyone knew. In October 2010, Thrun took to Google’s official blog to announce:
We have developed technology for cars that can drive themselves. Our automated cars, manned by trained operators, just drove from our Mountain View campus to our Santa Monica office and on to Hollywood Boulevard [more than 350 miles]. They’ve driven down Lombard Street, crossed the Golden Gate bridge, navigated the Pacific Coast Highway, and even made it all the way around Lake Tahoe. All in all, our self-driving cars have logged over 140,000 miles.11
That’s greater than the lifetime of an average car. By the summer of 2013, the cars had driven themselves more than six hundred thousand miles, virtually without incident. “Our work is focused on building cars that can drive anywhere by themselves, any street in California,” Thrun told the TED conference in 2011. Video shows Google’s car deftly dodging deer, joggers, tollbooths, people crossing the street, and bad drivers. The car had advanced so far beyond the 13.7 mph average speed of the DARPA Urban Challenge that other drivers hardly took notice of it. And instead of needing human intervention every ten miles, Thrun told me, “now we are at fifty thousand miles, which is the equivalent of four years of human driving.”
Still, that’s not quite enough for Thrun, who wants to get the technology so perfect that human beings could sleep behind the wheel and cars could drive themselves with nobody inside. “I want to get to a point where we vastly exceed human safety,” he said. “To the point where we feel confident that we can drastically reduce traffic fatalities before we launch it.”
Now that Google, a search-engine company, has achieved the greatest breakthrough in automobiles since the internal combustion engine, essentially as a side project, the legacy car companies are racing to catch up. Toyota, Nissan, BMW, Audi, and Volvo have all demonstrated prototype self-driving cars of some capacity, although none seems to have matched Google’s performance. One thing is certain: now that Thrun and his fellow pioneers have created the technology, Americans are going to want it—from whatever manufacturer can provide it.
If we have self-driving cars, how will we use them? If the millions of Americans who spend nearly an hour each day commuting to and from work have anything to say about it, we’ll probably want to start by taking a bite out of traffic congestion.
“Congestion” is a relative term. It might be hard to believe, but at the height of rush hour, vehicles occupy only 6 percent of the space on our highways. Think of an egg carton with only one egg in it. But with human drivers, that’s as close as cars can safely travel. Each year, highway use increases by about 3 percent, so the roads are nearing their limit, and we’re stuck in gridlock.
If cars could be packed more tightly on our roads, Thrun points out, we could double or even triple highway capacity. Fitting three cars into the space that now holds one would be the equivalent of adding four lanes to a two-lane highway. With self-driving cars, we could achieve that density, since they are constantly alert and they can react instantaneously. By the time they reach the market, moreover, they’ll probably be able to talk to each other—pinging the cars behind them with notices to slow down or speed up. The SARTRE project in Europe, an effort that includes Volvo and Ricardo UK as partners, has already demonstrated a “convoy” system of car-to-car communication that allows a lead truck to direct a train of cars following closely behind. (The configuration also yields substantial fuel savings—up to 21 percent—because of the decreased wind-resistance.)
Just as self-driving technology would let cars drive closer together, it would also allow them to travel much faster. Greater density at greater speeds could virtually eliminate the problem of traffic, even with no new roads. (The question remains how we can achieve these benefits fully if highways have a mix of traditional cars and self-driving cars. The solution might be roads reserved for self-driving vehicles, just as many states now reserve certain lanes or highways at peak hours for “high-occupancy vehicles.”)
Self-driving cars could do something even more important than eliminating traffic congestion—they could be dramatically safer. If they eliminated the overwhelming majority of accidents, we could make them out of lighter materials and substantially improve fuel economy. Self-driving cars could transform the lives of elderly and disabled persons, as well, giving them previously undreamed-of independence. Drunk driving would become a thing of the past.
Self-driving technology would revolutionize the shipping industry—businesses like UPS and FedEx, as well as long-haul trucking. Many shipping routes would no longer require drivers, and they could operate around the clock.
In 2013 I visited Caterpillar, which has been working with the Robotics Institute at Carnegie Mellon University on a self-driving truck for mining and construction sites. These are huge industrial devices, two stories tall and carrying loads of four hundred to five hundred tons. In an open-pit or strip mine, a self-driving truck could work seven days a week, twenty-four hours a day apart from time for routine maintenance. In some remote, high-cost areas, a self-driving truck could save up to a million dollars a year. Caterpillar is not alone in the quest for self-driving trucks. Rio Tinto is buying 150 self-driving trucks from Komatsu. Oshkosh Defense is building the TerraMax self-driving truck for the Defense Department. These developments are happening now.
The taxi business, and any business that relies on professional drivers, will face major disruptions. (This is one more example of why we will need a 24/7 online adult learning system for people displaced by technological change to learn new skills and new capabilities.)
It might be less obvious that adding self-driving capability to cars could similarly disrupt the automotive manufacturing industry. But the big automakers could be looking at a fundamental challenge to their business model. Today, the average American family owns two, maybe even three, cars. If your family is typical, every one of your cars sits idle in the driveway or in a parking lot twenty-two hours or more a day. That’s a lot of downtime for some very expensive equipment, but it’s often necessary because mom needs to drive to work, dad needs to drive to work, and the teenage kids need to drive to school and their babysitting jobs.
When families have genuinely self-driving cars, however, they may find that one is enough for the whole family. It ferries the kids to school at 7:30 while mom and dad get ready for wor
k. It returns home by 8:00, departs with mom and dad at 8:30, drops dad off at work by 8:45, and mom by 9:00. If dad has a lunch meeting downtown, the car is waiting for him by noon, ready to shuttle him into the city. The commute home is fast and leisurely, and both parents are enjoying a quiet dinner by 7:00, while the car runs to pick up the kids from soccer practice.
If vehicles can get from place to place on their own, they can spend more time doing what they’re made for and less time sitting in the driveway. But if millions of families that need three cars today will soon only need one, it will be a big hit to the auto industry’s bottom line.
That might only be the beginning of the disruption, however, because if cars can really move autonomously from point to point, there’s no reason for them to have any downtime at all. Maybe you will participate in a network of self-driving cars, like Zipcars that drive themselves, and simply pay as you go. When you need to make a trip to the grocery store, you will summon a car using your smartphone, and in minutes it will roll up outside, ready to take you. You will leave for work every morning at 8:00, and a car will always be there waiting for you. And because there are no drivers to pay and the entire fleet can stay busy constantly, the cost of such a service could be extraordinarily low.
In fact, participating in a shared self-driving car fleet would likely be a better choice for almost everyone than financing a $25,000 car. Lawrence Burns at Columbia University’s Earth Institute studied the potential for shared self-driving car systems and concluded that in a small city like Ann Arbor, Michigan, such a service would be 90 percent cheaper than car ownership, wait times for a vehicle would average about twelve seconds (two minutes at peak times), and only 15 percent of the city’s current number of cars would be needed. In a place like Manhattan, Burns concluded that rides in a self-driving car fleet would cost about a dollar per trip, compared with $7.80 per trip in yellow cabs. The service could use the smallest vehicle necessary for a particular journey: if you’re riding uptown by yourself, it’s much more efficient to take a two-seater.12 Who would have thought the personalization of mass transit could make it cheaper, too?
For longer trips, bus service might become more personalized and flexible. If twenty people in your neighborhood were going to New York today, along with five people in the next town over, the self-driving bus could pick you up at your house.
With new apps, you could send your own car to pick up dry cleaning or get take-out food, to refuel or recharge in its free time, to find the nearest open parking spot and wait there, or to pick you up for appointments on your calendar. A commenter on my Facebook page imagines parental controls that will let her children use the car for approved destinations. Pizza delivery, school buses, police cars, ambulances, and trash collectors could all benefit from customized artificial intelligence in self-driving vehicles.
The widespread availability of cars dramatically changed the American landscape following World War II. Entire towns grew up dependent on the automobile but offering affordable family homes with yards and one-car garages. Suburbs, drive-in movies, McDonald’s, A&W Root Beer, and Disneyland owed their success, or even their existence, to the automobile. The car became a symbol of freedom—a liberating machine for tens of millions of Americans. People could live in the country and work in the city. They could haul around their kids and their kids’ friends. They could take weekend jaunts to the beach. They could pull trailers and take their families to the Rocky Mountains. The car made life richer. As Thrun told me, “I believe the car as an invention had a bigger impact on the twentieth-century society than any other invention—I think more so than television—cities look different, work looks different, social relationships look different, almost everything has been impacted by the car.”
Self-driving cars have the potential to accomplish a transformation of the same magnitude. They will create unimaginable new opportunities for work and leisure, further shrink our sense of distance, change our landscape, and free up the useless hours we currently spend driving a car or sitting in traffic.
Imagine that you could roll out of bed at 5:30 in the morning and climb into the self-driving car that had just pulled up to your home. You could drift back to sleep as the car whisked you to work in the city a hundred and fifty miles away (less than an hour-and-a-half trip at 120 mph), waking just as you arrived around 7:00. Imagine that you could work throughout your evening commute, opening up time at home to spend with your family.
This scenario sounds like something from the life of a chauffeured celebrity or a Fortune 500 CEO. Some of it sounds impossible even for the superrich. But millions of average Americans could soon be living in such a world, finally liberated from the excruciating commute they once made every day.
All of this could happen. The technology Sebastian Thrun and others are pioneering could allow us to break out in transportation. Google engineers today are commuting to work in self-driving cars. Theoretically we could all start using them by 2020. But that almost certainly won’t happen. The technological challenge of designing Priuses that drive themselves pales in comparison with the legal slog it will take to clear the way for their use. We can work our way through the thousands of regulations that stand in the way of the breakout in transportation, but if we can’t overcome the prison guards who will emerge to delay the disruptive technology, it could be decades before we fully enjoy the breakthroughs that are here today.
The first and most obvious barrier is liability law. Who will be responsible if a self-driving car gets into an accident? The owner? The passengers? The manufacturer? A software vendor? What if the owner knew of a faulty sensor but did nothing? There are hundreds of hard cases that today’s liability system is poorly equipped to handle.
The only states that have legalized self-driving cars formally—Nevada, California, and Florida—still require a human operator to remain at the wheel, attentive and ready to take over at an instant’s notice. The laws don’t even allow the person in the driver’s seat to read while the car does the work. Unmanned cars are out of the question right now. Depending on the state, these requirements reduce the technology to a fancy form of cruise control.
Even if we could work out the complicated liability problems, hundreds of other federal, state, and local laws could stifle a real breakout. In many states, for instance, dense convoys of self-driving cars would violate tailgating laws. In addition to American laws, there are international laws to deal with, like the Geneva Convention on Road Traffic. All of these regulations alone could delay the availability of fully self-driving cars for decades.
If the challenge were simply the bureaucratic headache of sifting through outdated laws and rewriting them, at least we’d know what we had to do. But the legacy automakers and dealers, the Teamsters, the transit unions, the taxi drivers, the rental car companies, local law enforcement (which profits from traffic tickets), and others will not give up without a fight. Many of these prison guards, powerful interests whom self-driving technology threatens, will undoubtedly try to use the thicket of existing laws and regulations—and to erect new ones—to keep us trapped in the past. A struggle taking place today in the taxi industry is a preview of the fights to come.
Uber and Out
Ridah Ben-Amara was sitting in his Lincoln Town Car in downtown Washington, D.C., just before 9:00 on the crisp morning of January 12, 2012, when the iPhone on his dashboard buzzed to get his attention. Ridah was a livery driver filling his spare hours by picking up jobs for the fast-growing company Uber, which allows customers to summon black-car service using their smartphones and charge the trip to their accounts, no cash exchanged and no tip required. The iPhone’s GPS dropped a marker on the map where his customer—labeled “Ron”—was waiting. As Ridah approached the pick-up location a few minutes later, his phone called Ron to tell him his Uber ride was arriving.
Ron got into the backseat and directed Ridah to the Mayflower Hotel on Connecticut Avenue, a popular breakfast spot for the city’s elite. Ridah made his way t
hrough the capital’s infamous rush-hour traffic and pulled up in front of the Mayflower. Ron got out of the car quickly and was gone. The fare would be charged automatically to his credit card, so there was no money to exchange. It was a typically seamless Uber transaction.
Until officials from the District of Columbia Taxicab Commission surrounded Ridah’s car, demanding to see his taxi license and proof of insurance. “Ron,” it turned out, was Ron Linton, the D.C. taxi commissioner. Washington’s taxi companies weren’t happy about Uber’s new service, which was quickly gaining popularity and luring customers away from their dingier, less reliable, and cash-only cabs. Linton, after publicly declaring Uber to be illegal, took it upon himself to launch a “sting” operation against the company—which was operating openly and, it said, within the law.13
As the Washington Post reported, Linton’s DCTC goons proceeded to slap Ridah with $1,650 in fines for four supposed violations: “Not holding a chauffeur license, driving an unlicensed [taxi] vehicle, not having proof of insurance and charging an improper fare.” The Post quoted Linton’s accusation that Uber’s “improper fare” violated “a city law that says limousine trips must have a fare set in advance; Uber’s system uses time-and-distance metering.… Linton said Ben-Amara refused to cite him a fare before the trip began. ‘What they’re trying to do is be both a taxi and a limousine,’ Linton said. ‘Under the way the law is written, it just can’t be done.’”14 It just can’t be done. That’s an epigraph that should be carved over the entrance of every bureaucratic prison of the past in the country.
Linton followed classic prison guard tactics in using his interview with the city’s paper of record to warn potential Uber drivers that they could be next: “It kind of serves as a message to the others that are doing this, they’re not going to be immune,” he warned.
Breakout: Pioneers of the Future, Prison Guards of the Past, and the Epic Battle That Will Decide America's Fate Page 13