Lawrence D. Burns is a former vice-president of research and development at General Motors and a long-time adviser to Waymo, the Google self-driving car project. He is the author, along with Christopher Shulgan, of Autonomy: The Quest to Build the Driverless Car – And How It Will Reshape Our World.
These past few weeks have seen the Canadian automotive industry threatened as trade negotiations progress with my country, the United States. Those who have followed the talks have been struck by the way they’ve reflected aspects of two national identities: On the one side is President Donald Trump and his team – pugnacious, belligerent, the picture of capricious emotion; on the other is Canada and its Minister of Foreign Affairs, Chrystia Freeland – calm yet steadfast, all understated civility as she charts her cautious and sensible course.
What’s remarkable to me is the way this same duality played out during another drama that seems likely to disrupt the auto industry, albeit over the next decade. That drama is the development of the self-driving car – and there, the Canadian side’s national character served to ensure the project’s success.
Aside from the tortuous trade negotiations, that success is currently the auto industry’s most important talking point, creating unlikely bedfellows and spurring enormous investments. Uber, for example, announced just last week that it would invest as much as US$200-million to more than double the size, to more than 500 employees, of its self-driving car lab in Toronto – news that came on the heels of Toyota pledging to invest US$500-million to help Uber develop autonomous technology.
The commitments were only the latest in a series of billion-dollar announcements this year, as automakers such as GM and Ford and tech companies such as Apple race to catch up with sector front-runner Waymo, the company spun off from the Google self-driving car project in December, 2016.
I’ve argued for years that self-driving technology has the potential to transform the way we get around. Automobile transportation currently sees most of us relying on gas-powered vehicles that we own ourselves. But this system features an enormous amount of waste. We buy fast SUVs that can ferry five or more people for hundreds of kilometres, then let them sit in our driveways or parking lots 95 per cent of the time. And the hour each day that we do use them is dominated by short, low-speed drives with one or two people.
The recent investments made by automakers, ride-sharing companies and tech firms are designed to eliminate that waste from the transportation system. Rather than individual drivers owning their automobiles, my colleagues and I envision roads populated by self-driving, electric-powered robots, summoned at the touch of an app and paid for by subscription or on a per-trip basis. The new system would allow people to move about their environments for a small fraction of what it costs today. It would provide greater personal mobility to youth, people with disabilities and the aged. And it would eliminate more than 90 per cent of the approximately 1.3 million deaths that happen on the world’s roads each year.
What wrested self-driving technology from university-lab curiosity to something that could conceivably exist in the real world was the Google self-driving car team, known within the search giant by the code name Chauffeur.
Chauffeur’s remarkable safety record has been key to the growing public acceptance of self-driving technology. Consider that Chauffeur was in operation, and testing on public roads, for more than seven years before any of its autonomous vehicles were found to be at fault in a crash. And the term “crash” seems too serious to describe what happened. In February, 2016, one of the team’s self-driving Lexus SUVs bumped up against the side of a transit bus. It happened at just three kilometres an hour. No one was injured. The only evidence that something had happened was some scraped paint and dents.
Even today, after nine million miles of autonomous driving on public roads, none of the Waymo vehicles has ever been involved in a serious collision, let alone a fatal crash. Compare that with Tesla, which in May, 2016, saw former U.S. Navy SEAL Joshua Brown killed when his Autopilot-activated Model S crashed into the side of a tractor-trailer in Florida – just seven months after the public release of the Autopilot suite of self-driving features.
So who deserves the credit for Waymo’s remarkable safety record? Well, that’s the interesting part. During its first seven years of existence, the team was dominated by two personalities – one Canadian, the other American – each reflecting their national characters.
Both men built their reputations in robot-car races with million-dollar purses staged in 2004, 2005 and 2007 by the U.S. government’s Defense Advanced Research Projects Agency (DARPA), an obscure arm of the military with a mandate to develop distant technologies.
The American was Anthony Levandowski, the son of an American businessman father and a French diplomat mother, raised in Northern California and educated at the University of California, Berkeley, where he first became known as a brilliant, audacious engineer. Among the projects that earned him that reputation was the development of the world’s first self-driving motorcycle, which he entered into the first of the DARPA Grand Challenges – where it promptly tipped over mere feet from the starting line because he had forgotten to activate the gyroscopes that held it upright.
The Canadian, Chris Urmson, got his start as the technical lead of the Carnegie Mellon University teams that competed in the DARPA Grand Challenges, ultimately winning the last and most complex event. But it was in his role as chief engineer on the Google self-driving car team that Mr. Urmson’s characteristically Canadian qualities were so crucial.
He deserves a lot of credit for that project’s remarkable safety record. This becomes all the more notable when one considers the pressure he was under to roll out the technology as a commercial project. It was a peculiar dilemma: It’s believed autonomous technology will one day eliminate most of the annual fatalities on the world’s roadways. Yet rolling out the technology too soon, before it was safe to do so, seemed likely to cause fatal accidents, slowing the realization of the full safety potential of self-driving cars.
Some of the pressure also came from Mr. Levandowski, who lobbied the executives overseeing the project, such as Google X director Sebastian Thrun and Google co-founder Larry Page, to release self-driving products to the market well before Mr. Urmson thought the technology was ready.
Mr. Urmson was born in 1976, the oldest of three sons of Paul and Susan Urmson, an English couple who had immigrated to Canada because they thought it would offer better opportunities for their children. Paul worked for Correctional Service Canada, rising through the federal bureaucracy until he became the warden of Kent Institution, a maximum-security prison in British Columbia, and eventually assistant deputy commissioner of the CSC’s entire Prairie region.
The family moved a lot because of Paul’s work. Trenton, Ont. Victoria. Winnipeg. Finally, they settled in Saskatoon, which may be one of the most risk-averse and eminently Canadian places in an already risk-averse country. The point? Mr. Urmson grew up among people who viewed with suspicion those who drew attention to themselves. The guy is solid. Straight-shooting. Steady.
At a young age, his teachers assessed him as gifted, which qualified him to attend special classes with similarly intelligent children. Gifted teachers encouraged their students to enter a series of events then known as the Olympics of the Mind, which challenged participants to solve unconventional problems. Bit by the science and engineering bug, Chris also competed in science fairs. He would go on to study computer engineering at the University of Manitoba, where he spotted a poster outside his department office. “Come be a part of the robot revolution!” it read, with information about attending Carnegie Mellon University in Pittsburgh.
It was at Carnegie Mellon that Mr. Urmson would begin working for engineering professor Red Whittaker on the university’s teams for the DARPA Grand Challenges. Mr. Urmson wouldn’t win the first competition, although the vehicle his team created did travel the longest distance. Nor would his team win the second one. That honour went to a team led by Mr. Thrun, then a Stanford computer-science professor. But Mr. Urmson’s team did win the most complex of the events, the DARPA Urban Challenge in 2007, which led to an enticing job offer from his rival, Mr. Thrun, to become chief engineer of the Google self-driving car team.
Google’s Mr. Page and co-founder Sergey Brin motivated the team with an unconventional challenge: Build a robot car within two years that could navigate 10 of the hardest 100-mile drives on California’s public roads, and the founders would reward the team members with more money than they’d ever imagined. Under Mr. Thrun’s oversight, Mr. Urmson found himself leading a supergroup of robotics greats working feverishly against time on a secret project to achieve a goal set by two of the world’s richest men. And they did it. They met the Google founders’ challenge – with three months to spare before the two-year deadline.
Mr. Levandowski was also on the team. Mr. Urmson worked well with him during the project’s first two years. It was in 2011, after the achievement of the 10-difficult-drives challenge, that the tension between them became serious. Mr. Urmson wanted to keep the team’s engineering talent focused on developing ultrasafe self-driving capability, the so-called Level 4 and 5 that would see the vehicle’s software able to drive without human intervention. Meanwhile, Mr. Levandowski wanted to get a product to market – and fast.
The tension culminated in a battle for leadership of the team. “Anthony threatens to leave the team if he isn’t the single leader,” Mr. Thrun wrote in an e-mail to Mr. Page. “If he is the single leader, a good number of team members will leave.”
At that point, no one on Earth knew more than Mr. Urmson and Mr. Levandowski about how to make autonomous vehicles. Both men realized they stood to bring self-driving cars to market more quickly together than apart. But they found it incredibly difficult to work together. “It was pretty bad,” Mr. Page would say in court documents, years later, referring to what he calls the “fractious relationship” between Mr. Urmson and Mr. Levandowski. “I think they had a really hard time getting along, and yet they worked together for a long time. And it was a constant – yeah, constant management headache to help them get through that.”
Mr. Urmson and Mr. Levandowski were just such different people. Mr. Levandowski operated at a speed I’d never seen another human match. When he applied himself to solving a problem, his productivity was extraordinary. But the disadvantage of this character trait was that he would grow impatient when things didn’t progress at the tempo he thought possible. And if getting things done quickly required cutting a corner here or there, well, those issues could always be cleaned up later.
In contrast, Mr. Urmson was more meticulous. The engineer was focused on safety. Numerous times we spoke about the benefits that autonomous vehicles could bring to U.S. roads. The farmer’s math was compelling: Those 1.3 million roadway fatalities a year worldwide worked out to more than 3,500 deaths a day. The project Mr. Urmson was leading could save a lot of lives. And if the intent was to save lives, he was extremely aware that rolling it out before it was ready could be ruinous.
If Mr. Levandowski’s instinct was to just get the tech out there, to roll it out as fast as possible, Mr. Urmson’s instinct went in exactly the opposite direction. He devoted Chauffeur’s resources to whipping up a simulation team that would run the software in virtual reality, confronting the program with thousands of permutations of events, just to see what might trip things up and making corrections when a problem arose. Similarly, he set up a real-world version of these simulations at a 1,400-acre decommissioned military facility, the former Castle Air Force Base, about 160 km east of Mountain View, Calif. There, dozens of people, known as the Orange Team, devised scenarios that might bug up the self-driving software in the real world. For example, as one of the vehicles approached an intersection, an Orange Team member carrying a wall-sized canvas might step onto the street. Would the software recognize the form as a pedestrian or some other object, like a truck? What about a cyclist who happened to be carrying an enormous net bag of beach balls – would the car be able to understand that the entity would behave like a cyclist despite the unusual silhouette?
Such steps required time and money. They struck me as exactly the right moves. And they illustrated a rigorous approach on Mr. Urmson’s part that drove Mr. Levandowski crazy.
Mr. Levandowski left the Chauffeur team in January, 2016, complaining about, among other things, “too much BS” with Mr. Urmson and others. Then-Uber CEO Travis Kalanick soon bought Mr. Levandowski’s autonomous-truck startup, Otto, and installed the former Chauffeur engineer as leader of the ride-sharing giant’s self-driving car team.
What followed were a series of remarkable missteps. Uber began testing driverless vehicles in San Francisco without a permit from the State of California, earning public criticism when one of its SUVs ran a red light on the first day. Waymo ended up suing Uber after it discovered Mr. Levandowski had downloaded 14,000 files from its corporate servers the month before he left the company. Uber would subsequently fire Mr. Levandowski in May, 2017.
On the night of March 18, 2018, an Uber-operated Volvo XC90 SUV in autonomous mode collided with and killed 49-year-old Elaine Herzberg – the first time an autonomous vehicle was implicated in a pedestrian fatality.
In contrast, the safety record that Mr. Urmson engendered at Waymo encouraged one Morgan Stanley analyst to speculate this summer that the company could soon be worth US$175-billion – by far the biggest valuation for any firm in the sector. Mr. Urmson would leave the Chauffeur project in August, 2016, to launch his own startup. Known as Aurora and located in Palo Alto, Calif., the company would arrange a collaborative deal with Volkswagen and Hyundai and announce in 2018 that it had secured US$90-million in funding.
Aurora’s values statement reads like Mr. Urmson’s defiant response to spending so many years working with Mr. Levandowski: “We do the right thing even if it delays us or makes us less money,” it says. “No jerks: We debate and solve hard technical problems. We don’t waste time battling over personalities and egos and we have no tolerance for time-wasters and nonsense.”
In the quest to develop self-driving cars, the Canadian approach – risk-averse and meticulous – proved the one that ultimately created more safety and value over time. Whether a similar result will emerge from the NAFTA negotiations remains to be seen.