They’re a daily sight in their South of Market neighborhood, to the point that smartphone-obsessed pedestrians rarely even notice them. Passersby also may not realize that much of the time, the cars are driving themselves.
These cars, equal parts Detroit and Silicon Valley, are the creation of a San Francisco startup, Cruise Automation, and General Motors, which bought Cruise last year for well over half a billion dollars. They are just one component of a burgeoning autonomous vehicle industry that has developed with startling speed, one that could, very soon, fundamentally reshape our economy, our cities and our society.
For close to a century, vehicles that could drive themselves belonged to a future that forever seemed a few decades away, a jet-pack dream just out of reach. Now the technology is advancing fast enough that many of those developing it predict human-driven cars will be obsolete within a generation.
Ask any of them whether today’s toddlers will ever obtain a driver’s license, and the answer is almost unanimous.
“Not a chance,” said Cruise co-founder and CEO Kyle Vogt.
The creative engine for this world-shifting technology, its main laboratory and testing ground, is the Bay Area. More than 60 companies in the region — some based here, others drawn from as far away as China, France and Japan — are chasing the dream of robotic driving.
They include some of Silicon Valley’s biggest names: Tesla, Apple and, of course, Google, the Internet search giant whose willingness to throw its vast resources into self-driving research over the past decade forced skeptical automakers to follow.
Classic five-person startups bunking in garages also are angling for pieces of a market that all assume will be huge. Global automakers including Ford and Honda have opened Bay Area labs and placed billion-dollar bets on the field’s most promising technologies.
To the men and women engaged in this race, the prize is classic Silicon Valley: a chance to reap untold riches and change the world.
Self-driving cars, they say, will save many lives and transform others. They will prevent accidents, ease traffic and bring newfound mobility to the elderly and the disabled. They could also throw millions of cabbies, truckers, delivery people and others who drive for a living out of work.
There will be disruption, those in the industry say. But it will be worth it.
“I can’t wait, honestly,” said Sebastian Thrun, renowned roboticist and former head of Google’s self-driving car project. “It’s going to be such a great technology. Think of the 1.2 million lives we lose each year (to car crashes), mostly because they’re not paying attention. Think if we could get some of those lives back.”
Just how big will the market be?
Worldwide revenue for major automakers have reached the multitrillion-dollar level. Forecasts for self-driving vehicles cover a range, up to $87 billion in 2030 for sales of both autonomous and semiautonomous cars. But that’s just a start. The global “passenger economy” — including all the services, such as robot taxis and robot deliveries, related to autonomous vehicles — could hit $800 billion by 2025 and a jaw-dropping $7 trillion by 2050, according to a study from Strategy Analytics and Intel.
The stakes are high enough that some companies in this still-nascent industry have dragged each other into court, accusing rivals of breaking the rules to gain an unfair edge.
Witness the legal fight between Google’s self-driving division, now known as Waymo, and ride-hailing giant Uber, whose autonomous-vehicle efforts were until recently led by a former Google engineer. Waymo sued in February, saying that Uber and the engineer conspired to steal trade secrets. If a jury agrees, Uber’s program — and perhaps its survival — could be at risk. Tesla, too, briefly sought to hobble a competitor in court.
While the growing reality of an autonomous future is new, the vision is not.
Detroit’s big automakers have tinkered with self-driving cars for decades, bragging about the concept in the occasional magazine ad or at world’s fair exhibits. German engineers in the 1980s and 1990s equipped Mercedes-Benz vans and cars with cameras and other sensors, eventually letting one loose on the Autobahn, albeit with a human driver inside. At the same time, Humvees outfitted with robotic brains were lumbering around the campus of Carnegie Mellon University, Thrun’s alma mater in Pittsburgh.
But only within the past few years has the technology taken on a momentum of its own, even a sense of inevitability. Today, few in the industry doubt that this revolution is coming — soon.
The headquarters of Cruise Automation could pass for any other SoMa startup, at least on its main floor. Engineers sharing tables peer into monitors and wolf snacks from a well-stocked kitchen, beneath the benevolent gaze of a life-size pink stuffed ape.
The basement garage, however, tells a very different story. Rows of Chevy Bolts with sensors protruding from the bumpers and lidar — the laser equivalent of radar — on the roofs await their next spin on the streets. A 3-D printer fabricates parts as needed. Drivers confer over clipboards about the latest test results.
Vogt, 31, started Cruise in 2013 with friend Daniel Kan, 30, after selling streaming service Twitch to Amazon for $970 million. After that, Vogt could have done anything — or nothing at all. A confident engineer with a sly smile, he chose autonomous vehicles in part because he liked the technical challenge. He also saw the potential to create something truly transformative.
“I strongly believe this is the highest impact per unit of time I can possibly achieve as a person and an engineer,” Vogt said.
Cruise has quadrupled in size since GM bought it, straining the limits of its aging brick headquarters. GM and Cruise plan to invest $14 million to expand its operations, adding 1,163 full-time employees by 2021, at an average salary of $116,000. If those plans hold, Cruise could vault into the ranks of San Francisco’s 10 largest tech employers.
“It takes an army for a technical challenge like this,” Vogt said.
For GM, the Cruise purchase represents a bid to stay relevant as the industry it helped create experiences enormous changes.
GM has built cars since 1908 and sells millions of them each year. Faced with a new technology that could fundamentally alter how cars work — and how people use them — GM was determined not to get left in the dust by Silicon Valley upstarts.
Cruise now forms the core of GM’s self-driving efforts. This spring, GM forged an alliance with San Francisco ride service Lyft to employ the technology.
“Running our autonomous vehicle program as a startup is giving us the speed we need to continue to stay at the forefront,” GM CEO Mary Barra said in a statement about its San Francisco expansion.
For years, big automakers treated autonomous vehicle research as a sideshow — interesting, potentially useful someday, but nothing they needed to worry about. Many dabbled in early experiments. GM worked with the Radio Corporation of America in the 1950s on a system that could control cars through sensors and wires embedded in roadways. European automakers in the late 1980s and 1990s joined a project called Prometheus that included a 1995 drive between Munich and Copenhagen in a partially autonomous Mercedes sedan.
But given the limitations of the sensors and computers available, the big automakers never seriously invested.
“I can tell you that as late as 2014, I would meet with executives at car companies, and they would just smile at me, because it was known that this was crazy, and had nothing to do with reality,” said Thrun, Google’s former self-driving guru.
Today, virtually all of the world’s major automakers have jumped in. Ford, for example, has opened a bustling Palo Alto lab a mile from Tesla headquarters and has promised to have robotic taxis by 2021. And it will invest $1 billion over the next five years in Argo AI, an autonomous vehicle startup in Pittsburgh.
So what changed their minds?
In 2008, Thrun called a young Canadian roboticist named Chris Urmson to offer him a job at Google. The two had known each other for years, as colleagues and competitors.
They overlapped at Carnegie Mellon starting in the late 1990s. Long an engineering powerhouse, the school by then had become one of the world’s premier hubs for robotics research. Thrun, a native of Solingen, Germany, co-directed the university’s Robot Learning Laboratory. Urmson, a computer scientist, sought out its graduate program after seeing a poster of a Carnegie Mellon robot crawling out of a volcano.
But only after Thrun left for Stanford University in 2003 did the two get to know each other well, thanks to an event Urmson calls “the Woodstock of robotics.”
Sponsored by the Defense Advanced Research Projects Agency, the deep research wing of the U.S. military, the DARPA Grand Challenge dared teams of engineers to build vehicles that could drive themselves across more than 100 miles of desert terrain. Many of the participants weren’t sure it could be done.
Carnegie Mellon researchers had been designing early versions of self-driving vehicles since the 1980s, even taking one on a human-assisted cross-country road trip in 1995. For the first DARPA challenge in 2004, the university’s team assembled an autonomous Humvee named Sandstorm, with Urmson serving as technical director. But come race day, Sandstorm managed only 7 miles before getting stuck on a rocky berm. The rest of the field fared even worse.
“It went through, I think, three fence posts,” Urmson said. “You train for a marathon, and you get 2 miles into the 26-mile race — that’s kind of the equivalent of what we got.”
Thrun, by then the head of Stanford’s Artificial Intelligence Laboratory, didn’t compete in the 2004 race. But he saw the dismal results. Convinced that he could do better, he assembled a team.
DARPA scheduled its second race for Oct. 8, 2005, picking a site along Interstate 15 on the California-Nevada border. Carnegie Mellon was favored to win — if anyone did. But five hours into the race, a blue Volkswagen Touareg with a Stanford logo splashed across its windshield passed Carnegie Mellon’s latest Humvee. Dubbed Stanley by Thrun’s team, the Touareg finished the 132-mile course in six hours and 54 minutes, navigating every inch by itself.
At the time, few people outside academia, the government and the tech press noticed. But for the self-driving car, that day was its Kitty Hawk moment, the instant in which a long-sought technology finally proves possible.
“Before that, we didn’t think it was real,” Thrun said. “We didn’t really think you could take over from human drivers. Even if you have a car that can drive itself 97 percent of the time, that 3 percent is a lot.”
The contest also served as an incubator for the self-driving car industry. Many of the young engineers competing in the desert that day would, within just a few years, start the companies now pushing this dream closer to reality.
Google co-founder Larry Page had attended the race out of curiosity — showing up in a hat and glasses for disguise, according to Thrun — and the two men met. In 2007, Thrun joined Google. Page and his co-founder, Sergey Brin, saw in autonomous vehicles the potential for a remarkably disruptive technology that could save lives and reshape society. They gave Thrun a green light to hire.
He called Urmson and started stocking his team with other DARPA challenge veterans. They included Anthony Levandowski, whose UC Berkeley team entered a self-driving motorcycle dubbed Ghostrider, and Mike Montemerlo, one of the brains behind Stanford’s Stanley.
In 2009, Page issued Thrun’s team a challenge: devise cars that could drive 100,000 autonomous miles on public roads and navigate 10 courses along such well-known California routes as San Francisco’s twisting Lombard Street, several bridges across the bay and stoplight-infested El Camino Real.
Quietly, Toyota Priuses outfitted with radar, cameras and rooftop lidar that looked like a madly spinning Kentucky Fried Chicken bucket hit the streets. The cars began logging mile after autonomous mile in Silicon Valley and San Francisco, even following Highway 1 along the sheer cliffs south of Big Sur. Humans sat behind the wheel, ready to take control if needed.
The humans often found it nerve-racking. And at times, astonishing.
One day Urmson and software engineer Dmitri Dolgov were riding in a Prius around Lake Tahoe when the car abruptly slowed to a stop next to a line of parked cars. Urmson couldn’t tell why.
“I turned to Dmitri and said, ‘Why on earth are we stopping? This is dumb,’” Urmson said. “And then a little old lady walked in front of the car. It had seen her and was yielding to her in the crosswalk, and I couldn’t see her. That was cool.”
By late 2010, Google was ready to start talking publicly about its work. This, too, would prove to be a pivotal moment.
For decades, few people had paid much attention to autonomous vehicle research. Google’s cars — first the Prius, then the Lexus sport utility vehicle that followed it, and finally the company’s bubble-shape car built from scratch — changed that, seizing the public’s attention. The auto industry took notice.
After their near-death experiences during the 2008 recession, automakers had turned to Silicon Valley for new technology, opening labs in the Bay Area and hiring local engineers. Now, they realized that the advent of the driverless car wasn’t that far off — and someone else was perfecting it.
“All of a sudden, the world woke up,” Thrun said.
The Bay Area’s home-grown car company, Tesla, also took up the challenge.
Led by PayPal co-founder Elon Musk, Palo Alto’s Tesla had introduced a long-range luxury electric car before any of the traditional automakers. In September 2014, Tesla began equipping all Model S sedans built at its Fremont factory with the sensors they would need to steer themselves at freeway speed. In October 2015, the company sent those cars a wireless software upgrade activating the Autopilot system — not true self-driving technology, but a big, widely watched step in that direction.
With that, the race to develop autonomous vehicles abruptly shifted from a friendly academic competition to a flat-out sprint involving Google and Tesla, previously doubtful automakers and Silicon Valley startups. A profusion of companies developing software, hardware or entire vehicles quickly sprung up across the Bay Area.
Many shared DNA, which only made the competition more fierce.
Levandowski, the creator of Ghostrider, left Google to form Otto, which focused on self-driving trucks. Otto, in turn, was snapped up by Uber, whose co-founder, Travis Kalanick, said his company might not survive if someone else started making robot taxis first. Levandowski now stands at the center of the courtroom fight between Waymo and Uber, accused by Waymo of stealing designs for a better lidar.
After Levandowski refused to testify, Uber fired him in May. Kalanick, beset by scandals involving Uber’s corporate culture and attempts to thwart government regulators, stepped down as CEO in June. But none of that stopped the suit. The trial is scheduled to begin in October.
Urmson left Google and founded Aurora Innovation with Sterling Anderson, the former head of Tesla’s Autopilot program. Tesla sued them in January for allegedly stealing confidential information, but settled three months later.
Thrun decamped to focus on Udacity, his online technical education platform, which offers classes in programming self-driving cars. Auto industry veteran John Krafcik, the former CEO of Hyundai Motor America, now heads the Google program, which was renamed Waymo late last year.
Test vehicles roam the streets of San Francisco, Mountain View and nearby cities every day. Thirty-six companies have California permits to test autonomous vehicles on public roads.
Automakers from around the world, meanwhile, bring their robotic vehicles to GoMentum Station in Concord, a former navy munitions depot whose grid of roads and clusters of empty buildings can simulate a city or a stretch of highway. Closed to the public, GoMentum gives engineers space to tinker with their cars far from traffic or pedestrians, with no one but wild turkeys and prairie dogs looking on.
Venture capital, the lifeblood of Silicon Valley startups, has flowed into self-driving cars with increasing speed. Global investment totaled $577 million in 2016, according to PitchBook, which tracks VC spending. That’s 12 times 2014’s sum of $48 million.
Most of the money flooding in, though, comes from cash-rich companies — global automakers, Uber, Google parent company Alphabet, and Silicon Valley chipmakers like Intel and Nvidia.
Blockbuster deals abound. Uber shelled out $680 million for Otto. Intel is paying $15.3 billion for Israel’s Mobileye, which makes vision-based driving assistance systems.
Some companies — like Morgan Hill’s Velodyne, which builds the lidar that allows cars to sense distances to other objects — focus on hardware. But there’s a limited tool kit of equipment that nearly all self-driving vehicles use: lidar, radar, cameras and computers. What distinguishes one robot car from the next is code, the programming to interpret what the sensors see, integrate it with data from hyper-detailed 3-D maps, and based on that information, decide what to do next.
Building a car is an expensive proposition; writing code isn’t. That leaves an opening for startups.
Stepping out of noon Mass at Mission Santa Clara, Susanne Jung, 79, adjusted her flowered hat against the bright sun and spied an odd-looking car nearby. Signs on the boxy vehicle advertised “Free Campus Shuttle powered by Auro” and “Self-Driving Vehicle.”
“I have to try it today,” she said, clutching her purse and climbing into the back seat. “I see it here whenever I come to Mass, but usually I’m too nervous.”
Chugging along at a steady 7 mph, the shuttle glided down the palm-lined walkways of Santa Clara University, occasionally shuddering to a halt whenever a pedestrian crossed its path. Eric Weber, its backup driver, showed Jung that his hands weren’t on the controls as the car rolled up to a school building, paused at a shuttle stop, then made a U-turn to continue its mile-long route around campus. It rolled past oblivious skateboarders, groups of students and campus visitors, some of whom whipped out their phones to take photos.
“I can’t wait to tell my children and grandchildren,” Jung said. “I never dreamed of something like this.”
The Santa Clara campus lies about 10 miles from Waymo’s headquarters, and 13 miles from Tesla’s. But its shuttle — the only one of its kind in the state — was cobbled together by a tiny startup, Auro Robotics.
Auro’s three founders came to the Bay Area from India in 2015 to attend Y Combinator, the elite startup incubator that’s more selective than Harvard. Their idea: concentrate on low-speed autonomous shuttles for colleges, amusement parks, senior residences and corporate campuses — places where navigation would be relatively simple. Avoid public roads, reducing technical and regulatory hurdles.
They rented a four-bedroom ranch house in Sunnyvale, converted the garage into a workshop, found some unpaid interns and bunked together for six months, working around the clock and subsisting on takeout from a nearby Indian market.
“If you’re working on autonomous vehicles, you have to be in Silicon Valley,” said Auro CEO Nalin Gupta, 27. “The talent pool is concentrated here. The access to capital is here. And people here are much more receptive to new ideas.”
By January 2016, Gupta and co-founders Srinivas Reddy, 26, and Jit Ray Chowdhury, 30, had $2 million in seed funding and a tentative agreement with Santa Clara University to use its 110-acre campus as an open-air laboratory. The shuttle was up and running by November.
Now comprising 10 people and housed in an actual office, Auro is working on its shuttle 2.0, creating a boxlike roof attachment to hold most of the wiring and sensors. The attachment can be swapped into vehicles of different sizes.
Carving out a place for itself in a crowded field dominated by much larger and richer companies isn’t a big problem, said Ben Stinnett, 26, Auro’s chief operations officer, who has since left for a robotics startup. Auro has focused all its energies on its specific niche and is seeking deals with other universities, as well as Mexican resorts.
“It’s one thing to log test miles with an empty vehicle in the conditions you get to pick,” Stinnett said, “but it’s an entirely different thing to log those same miles with passengers in your care and customers you have to answer to, which is exactly what Auro is doing.”
Auro’s shuttle is currently the only opportunity in California for the general public to ride in a self-driving vehicle. But that will soon change.
French company EasyMile is scheduled to start ferrying workers around Bishop Ranch office park in San Ramon using bright red shuttles it’s been testing at GoMentum Station.
Uber’s brief test of robot taxis in San Francisco in December ended abruptly when regulators swooped in. It’s now back to testing in the city, but without passengers.
Although they clamped down on Uber, California officials are allowing most of these first, tentative steps to proceed, despite complaints from some consumer advocates that the technology is not yet safe enough for public use. For the state, it’s a delicate balancing act, guarding safety without piling on so many regulations that companies flee to other, more lenient locales. Phoenix, Pittsburgh and the state of Michigan all actively court autonomous vehicle research.
Stinnett has ample reason to hasten the arrival of driverless vehicles. His mother’s multiple sclerosis has worsened to the point where she can no longer walk or drive.
“My dream for her is neighborhood autonomous vehicles, so she’d have mobility,” he said.
Stinnett’s conviction is hardly unique. The engineers and entrepreneurs developing this technology share a fundamental belief that autonomous vehicles will save lives, provide mobility and independence, and boost economies worldwide.
Many also acknowledge the downside.
For one, autonomous vehicles could render millions of truckers, ride-share drivers, cabbies and delivery drivers obsolete. That enormous shift could begin soon; Ford, Lyft and others plan to have self-driving taxis in the next four years.
“There’s near-universal consensus that at least 4 million transportation workers will be displaced by this technology, and yet no one is talking about what is the plan for those workers,” said Doug Bloch, political director of the Teamsters Joint Council 7. “It’s going to be a crisis.”
Other potential drawbacks range from annoying to dystopian.
City streets full of constantly circling autonomous taxis could become perpetually crowded, even if traffic flows smoothly. Accidents might be less frequent but more deadly, as cars drive faster and closer together. Hackers could gain control of robot cars and trucks, turning them into weapons or using them to disrupt traffic. Terrorists could load a bomb into a self-driving car and send it to a target, like a low-speed cruise missile on wheels. Control centers tracking cars could invade the privacy of passengers.
These nightmare scenarios are far from certain. And yet, even analysts who think autonomous vehicles will do more good than harm say their arrival is likely to alter society in profound ways.
Car ownership will almost certainly decline, as it becomes cheaper and easier to simply hail a robot taxi. Within a couple of decades, driving one’s own car may become a hobby, like riding horses.
“People will be able to subscribe to transportation services just like they do to entertainment with Netflix or Spotify,” Lyft President John Zimmer said last year.
For a nation whose century-old love affair with automobiles reshaped its physical landscape — from interstate freeways to downtown business districts to suburbs with a garage in every home — giving up personal cars would be a seismic shift.
Robot cars will first roll out as fleets of self-driving taxis in controlled environments — the downtowns of congested cities, business parks, retirement communities, college campuses, dedicated freeway lanes. At first, those fleets will probably operate in dense cities with mild climates. Car sensors still struggle to see through snow and rain.
“San Francisco is one of the most likely U.S. cities, from a weather standpoint,” said Sam Abuelsamid, senior research analyst at Navigant Research. Other early hot spots could include Singapore, which has already extended a welcome to robot-taxi tests, and China.
There are compelling reasons most people won’t own self-driving cars.
Cost is one — early models may hit $100,000 each. Even with economies of scale, expect self-driving cars to cost at least $10,000 more than standard vehicles.
Liability is another. “Manufacturers will want to keep these vehicles within their control to make sure they are properly maintained, serviced and kept up to date,” Abuelsamid said. “When you buy a car today, once it’s off the lot, the customer can do anything with it.”
Once the technology is perfected, however, many things will begin to change, including the streets themselves.
Highways may see lanes dedicated to autonomous cars. San Jose officials already are considering a driverless lane leading to the city’s airport.
City centers may “geofence” downtown areas during certain hours to ban human drivers or charge them a toll, the way that London now slaps a congestion charge on vehicles in its central district during weekday working hours.
In time, a space-constrained city like San Francisco, served by a fleet of shared robot vehicles, would need fewer parking lots and garages. Those spaces could become cafes or mini parks. Large garages could be bulldozed to make way for condominiums and apartments. Home garages could become in-law units.
“Cities can be designed for people rather than vehicles,” said Nidhi Kalra, senior scientist and director of the Bay Area office for the Rand Corp.
For both good and ill, the autonomous future is coming. Onetime doubter Chris Urmson is sure of it.
“Self-driving cars have gone from complete science fiction to clearly already solved in near-record time,” he said. “And remember: This hasn’t actually happened yet, right?”
©2017 the San Francisco Chronicle Distributed by Tribune Content Agency, LLC.
