Sebastian Thrun
Innovators are unhappy, says Sebastian Thrun. They feel the pain of an unsolved problem and ask, "What can I do about it? And what can I get other people to do about it?" His pain is personal: When he was growing up in Germany, a car accident killed a childhood friend, and last November, another accident claimed the life of a coworker. Both crashes were avoidable, Thrun says. Which is why he's developing unmanned robotic cars that drive more safely. "I'm in service of humanity," he says.
In 2005, Thrun, then a Stanford professor, led the winning team in the government's $2 million DARPA challenge. His vehicle drove itself across the 132-mile desert racecourse. Afterward, he hired two of his competitors to join him at Google and take their work to the street. So far, their unmanned vehicles have driven more than 140,000 miles, along California's coast, through highway construction zones, local traffic, and even on San Francisco's famed Lombard Street. ("Safety drivers" sit behind the wheel but very rarely take control.) Each car is equipped with sensors that gather millions of data points per second about the surroundings and differentiate cars, pedestrians, birds, you name it. A radical idea? Not if you consider that onboard computers already land planes when pilots don't have enough visibility.
Thrun's breakthrough was looking at how humans learn to drive and applying it to computers. "You're not programmed with a list of 15,000 rules of what could happen," he says. "You drive, you ride with someone else driving, you experience it. So our cars experience it. We train them to learn and get better the more they drive."
More than a million people worldwide die every year in car crashes, most caused by human error. Because robotic cars don't get distracted, says Thrun, they can travel closer together, reducing stop-and-go traffic. Not only could they save lives, but they could save commuting time and fuel too. He's also enthralled by the prospect of reducing the number of cars on the road, now more than 1 billion. With the right technology, cars could be shared and used wherever they're needed, instead of being parked the majority of the time.
Thrun knows there are years of technical, regulatory, and legal obstacles ahead, but he's undaunted, partly because his cars will keep learning and improving. "There's almost no problem that can't be solved," Thrun says. "That's important as a premise. History has proven it over and over again."