These Truckers Work Alongside the Coders Trying to Eliminate Their Jobs
Just before Stefan Seltz-Axmacher offers a job to an engineer at Starsky Robotics Inc., a driverless trucking startup in San Francisco, he gives them the talk.
This is a company that employs truck drivers, is how the talk begins. The coders are sometimes taken aback—this differs from the usual change-the-world spiel deployed in hiring meetings. Truckers have very different ideas and different experiences from people like you, Seltz-Axmacher continues. Statistically speaking, many of them are Trump voters. They will say things that you may find startling. Not in a malicious way, but because people from, say, rural West Virginia talk differently than people from San Francisco. Can you handle that?
“Not everybody can,” Seltz-Axmacher says over beers in Fort Lauderdale, where Starsky does some of its testing. “And that’s OK.”
Most driverless vehicle operations, including those at Ford Motor Co. and Alphabet Inc.’s Waymo, are focused on developing cars or trucks that operate with no human oversight at all, or “level 4 autonomy.” The idea is that a passenger could safely take a nap, send a text, or tie one on while the software worries about the road, but that kind of freedom could be decades away. Seltz-Axmacher, Starsky’s co-founder and chief executive officer, who’s featured in this week’s Decrypted podcast, is attempting something that’s both more modest and, potentially, more disruptive to U.S. employment. His company has designed an artificial intelligence system for big-rig trucks that makes them mostly self-sufficient on highways, and then, when it’s time to exit onto local roads, allows them to be taken over and driven from a remote operations center. The plan is to eventually employ dozens of drivers, each of whom will keep an eye on a few trucks at once, sitting before arrays of monitors livestreaming views of windshields and mirrors. The company’s name is a reference to a CB radio slang term for when drivers work in teams—that is, like the title characters of the 1970s TV series Starsky & Hutch.
Most of Starsky’s AI rivals are focusing exclusively on research, logging as many miles and as much performance data as possible. Seltz-Axmacher’s trucks are still in beta, too—but they’re already earning revenue, carrying containers full of goods along U.S. highways. While the remote-control system develops, two Starsky employees ride in each cab: a software engineer in the passenger seat, keeping an eye on the algorithms, and a truck driver behind the wheel. This proximity is why there’s a second talk.
We hire truckers, Seltz-Axmacher tells prospective drivers right before offering them a job. But we also have a lot of engineers in Silicon Valley. Everything you’ve heard about San Francisco—it’s all basically true. There is something called raw denim, and in San Francisco people wear it, which means that some of your colleagues will pay up to $300 for a pair of blue jeans. They sometimes drink $7 lattes, too. Many of your co-workers will not be from the U.S. They will have accents. Can you handle that?
The drivers all say yes, but really, not everyone can. Since Starsky’s founding in 2015, Seltz-Axmacher has parted ways with two of the eight drivers he’s hired. One used an anti-gay slur to refer to a fellow driver, which worried Seltz-Axmacher because Starsky headquarters is on Folsom Street—home to the Folsom Street Fair, the famous leather festival held every September. “His first day was Wednesday, and his last day was Thursday,” Seltz-Axmacher says. “Mixing blue-collar workers with people who have postdocs is hard.”
Economically speaking—that is, in the most brutal terms—truckers are disposable. Almost anyone can become a professional driver with a month or so of training, and most don’t stick around for long; median pay is about $40,000 per year, and the work is often unhealthy, painful, and lonely. Software engineers, on the other hand, are some of the best-paid, hardest-to-hire employees in the modern economy. The variety that Seltz-Axmacher employs—specialists in AI and machine learning—are even better paid and even harder to hire. Google has been known to pay its self-driving car engineers millions or even tens of millions. Starsky’s coders don’t make that much, but the point remains: In its cabs, side by side, are representatives of some of the most and least promising careers in America.
Starsky’s offices have high ceilings and two dozen open-plan desks. It’s not fancy—the furniture is cheap, the carpets look old, and the coffee comes out of plastic pods—but the company’s engineers come from some of the world’s top research universities, including Carnegie Mellon, Stanford, and the University of California at Berkeley. Of the six truckers on staff, one or two are usually in San Francisco, and the rest are on the highway. “We basically have people from two worlds, neither of which has ever talked to each other,” says Seltz-Axmacher, who grew up in suburban Maryland. “That’s kind of what’s wrong with this country.” His hope is that Starsky, by employing truckers who oversee trucks from offices and work alongside engineers, can help bridge the divide.
Of course, Starsky is a for-profit business, not a truth and reconciliation commission. It’s one of a handful of companies trying to seize a piece of the trucking industry’s $700 billion in annual revenue. Starsky has raised $5 million in seed capital from, among others, Y Combinator, the Silicon Valley venture fund and incubator. Its competitors include Embark, which is also backed by Y Combinator, and Otto, a startup that raised no outside capital and had fewer than 100 employees when Uber Technologies Inc. acquired it for $700 million. (Otto is the subject of a lawsuit that claims its co-founder stole technology from Alphabet, Google’s parent.) A fourth company, Peloton Technology, has raised $78 million to pursue adding some autonomous capabilities to conventional trucks. There are also self-driving big-rig programs inside Alphabet, Tesla, Volvo, and Daimler. All of these companies want to avoid alarming truckers, their employers, and regulators. But if any of them succeed, they will drastically reshape the labor market in one of the country’s most important industries.
Three and a half million Americans drive trucks for a living, making it one of the most common jobs in America. The larger trucking economy—including cargo brokers, truck manufacturers, truck stop waitresses, and so on—accounts for an additional 4 million jobs, according to the American Trucking Associations (ATA), a trade group. A huge proportion of them are threatened by a decade of driverless research coming out of universities and Silicon Valley companies.
A truck traveling hundreds of miles to make a delivery represents an almost ideal application for the latest autonomous driving technologies. Long-haul truckers spend much of their time on interstate highways, where curves are gentle, lanes are well-defined, and pedestrians and bicycles—the bane of any AI vehicle engineer—are prohibited. Trucks are big and heavy, so they’re easier to outfit with special sensors needed to control them. All of this has caused trucking to be seen by automation experts, and in the popular press, as a test case for the impact of AI on employment. If a lot of long-haul truckers lose their jobs, then maybe lawyers and accountants—whose work is often repetitive—should be worried, too. But a major difference between trucking and those fields is that it’s a job few Americans seem to actually want to do.
Truck tonnage—the weight of freight carried—is up by more than 30 percent in the U.S. since 2009, according to the ATA, while the industry’s labor force has grown by about 10 percent. The trade group has estimated there are 48,000 open jobs, a figure that’s expected to more than triple over the next decade. “It’s just more and more demand on the industry, and fewer people coming into it to drive the trucks,” says Chris Spear, the ATA’s CEO.
Federal law limits truckers to 70 driving hours over eight consecutive days. But because drivers are paid by the mile rather than the hour, many fudge their time sheets to drive more hours. On a good day, an entry-level driver might make about $15 an hour. On a bad day—one spent in traffic or sitting in a port waiting for paperwork—he might make just a few dollars an hour. Drivers generally spend several weeks on the road at a time, sleeping in their cabs at rest areas. They gain weight and get lonely. The annualized turnover rate among drivers at large “truckload fleets” is 71 percent, the ATA says. “Most people who try it out decide, given the pay and the conditions, it’s not a very good job,” says Stephen Burks, an economist at the University of Minnesota at Morris and a former trucker himself. “People are voting with their feet.”
Many of us, when we think of trucking, don’t see the industry this way. We think of freedom and the open road. We think of Convoy, the novelty country song that made it to No. 1 on the Billboard pop chart in 1976, or of Smokey and the Bandit, which would have been the nation’s highest-grossing film of 1977 if not for Star Wars. We think of a job that’s necessary and steeped in Americana.
Whatever truth these ideas once possessed has faded. The union-friendly rules that once helped make trucking a well-paid blue-collar job were dismantled by a series of reforms, culminating when Jimmy Carter signed the Motor Carrier Act and deregulated the industry in 1980. Membership in the mighty Teamsters union plummeted, and the short, regular routes that allowed truckers to go home most nights were replaced by a system in which truckers are treated a lot like Uber drivers. “The amount that they’re getting paid per mile is really a small fraction of what they were getting,” says Michael Belzer, an economist at Wayne State University and a former driver who wrote a book on the industry called Sweatshops on Wheels. “It’s not an exaggeration at this point to suggest that it’s half the pay.”
Thanks in part to the advent of mundane technologies, such as automatic transmissions, that make driving easier to learn, the industry has moved away from employing career truckers and toward a model of paying little more than minimum wage and constantly replacing the drivers the industry churns out. Commercial licensing schools charge about $5,000 for a five-week course, but trucking companies will advance applicants the fees and then deduct the tuition from the new hire’s salary. “There’s a shortage, in part, because the industry wants it,” says Steve Viscelli, a sociologist at the University of Pennsylvania who studies the industry. “It’s cheaper and easier to manage the problem through high turnover.”
This has turned trucking into a kind of economic safety valve—work you do when you’re out of options. The industry puts a more positive spin on this. “There’s a lot of pride that goes into moving the nation’s freight,” says Spear. But in a 2015 video produced by the ATA, the group’s chief economist Bob Costello suggested lowering the interstate truck driving age, currently 21, as a way to better compete for young people who would otherwise choose military service. “Often it’s a job of last resort,” he acknowledged. In other words, it’s pretty much the opposite of being a coder.
Jeff Runions, a Starsky truck driver, is sitting in the front seat of Buster, a late-model Freightliner Cascadia that Starsky leases and has modded out with cameras and sensors. He glances at Kevin Keogh, an Irish-born AI specialist who previously worked at Jaguar Land Rover and who’s been tapping out a few last-minute adjustments to the Starsky code while Runions does the driving.
“Good when you are.”
Runions flips a blue switch on a little panel bolted onto the center console. “All right,” he says. “She’s hot.” Runions cautiously takes his hands off the wheel and slides his foot off the accelerator. We are just west of Fort Lauderdale, cruising up Florida’s Highway 27 on a windy morning in late May, with the Everglades stretching out on either side. Runions says we’re near a stretch of road truckers call “Alligator Alley,” and sure enough we soon see an enormous dead gator on the shoulder.
Starsky is testing in other states, but Florida is an attractive proving ground, because it’s especially relaxed about driverless vehicles. Unlike Nevada and California, for example, Florida doesn’t require a special permit to conduct tests on public roads, or any additional insurance, or even a human being behind the wheel, as long as a licensed driver is “operating” the vehicle by remote control. Nevada laws are written so the state could allow remote-control driving in the future; in Florida, any licensed driver can do it today without asking permission, which is exactly what Starsky intends to do later this year.
In the meantime, there are still lots of problems to solve—like wind. Not long into our drive, a gust hits our left side, and the truck lurches toward the shoulder; the wheel turns left, overcorrecting and sending us drifting into the next lane. The experience is terrifying, although Runions and Keogh seem unfazed.
“It’s got to adjust, that’s all,” Runions says, explaining that the combination of wind and weight—today’s load is 20 tons, more than in other tests—represents a novel challenge. He keeps his hand on the blue switch and his eyes on his side mirror to make sure we don’t cut off anyone. He looks tense, but the truck finds the right lane after a few seconds.
Keogh says everything is normal. Starsky’s software is written to determine how hard the wind is blowing, he says, and then to steer against the wind and stay in the lane. But early on in a session, the computer isn’t fully calibrated yet. Runions offers a comparison: “You know how you are in the morning before you have your coffee?”
A few minutes later, he and Keogh seem comfortable, cracking jokes about the size of the alligators near the farm where Keogh grew up in Ireland. At another point, Keogh says, “I think we’ve zoned in on the correct control parameters.”
“Have you, now?” Runions shoots back, and then adds, “I’m learning to speak Irish.”
The two men have a good rapport, but they couldn’t be more different. Keogh is 27, graduated from Dublin’s prestigious Trinity College, and got a master’s degree studying robotics. Runions, who has a shaved head and a salt-and-pepper goatee, is 58 and didn’t graduate high school. When asked how he got into long-haul trucking, he responds instantly. “White Line Fever,” he says, without taking his eyes off the road. “Watch that movie.” In the film, Carrol Jo “CJ” Hummer leads a strike against an abusive transportation conglomerate, Glass House, that culminates when CJ drives a bullet-riddled truck straight into corporate headquarters, a literal glass house, and is shot in the face.
Runions had a rough childhood. He grew up in the foster care system in Wrigleyville, in Chicago, back when it was a bad neighborhood. He partied a lot, got into trouble, and then, at 16, got kicked out of the house. “I’ve been on my own since then,” he says. He took a job doing construction work in Savannah, Ga., and eventually found his way to a job at Great Dane, the trailer manufacturer.
Runions started driving full time in 1979 and eventually married a truck driver—he and his wife, Marlene, met at a truck stop in Atlanta in 1999—and has had pretty much every job you can have in the business. “It sucks,” he says about life on the road. “I was gone 21 days a month. If you stayed here for a couple days, you’d know what I’m talking about.” He barely saw Marlene and put on 75 pounds.
Starsky pays its truckers about $55,000 per year and gives them benefits and stock. Runions, as the company’s top driver, earns more and has fairly sane hours. He sleeps in his own bed, in a small house outside of Jacksonville, most nights. “Some people are really negative” about driverless trucks, says Runions, who read about Starsky’s technology and applied for the job online. “Then you tell them they’re going to have 40 hours a week instead of being gone all the time. People think you’re taking their jobs, but you’re not.”
Seltz-Axmacher, who is watching us from the back of the cab, nods in agreement. He envisions climate-controlled “driver centers,” in towns like Jacksonville, where people like Runions will work regular shifts in front of computers, without the greasy food or loneliness that has traditionally gone along with being a trucker. Starsky, he believes, has “the ability to make 3.5 million people’s lives a lot better.”
Not everyone agrees, of course. In May, drivers for Airgas Inc., which distributes industrial gases, went on strike in part because of a proposed contract provision that could allow the company, a subsidiary of the Paris-based Air Liquide, to use autonomous trucks. And in New York City, the union-backed group New York Communities for Change is mounting a campaign to urge the federal Department of Transportation to cease all funding for autonomous vehicle research until a plan is put in place to protect any displaced drivers.
If Silicon Valley companies aren’t forced to consider what happens to today’s drivers, “we will all lose our jobs,” says Rolando Perdono, one of the activists. “We won’t have anything to hold on to.” Perdono, 45, was born in the Dominican Republic. His English isn’t great, he didn’t make it through high school, and he has five kids to support. He’s been behind a wheel since he came to the U.S. 16 years ago and currently works as a local delivery driver for a cleaning-supply company. Perdono doesn’t love what he does for a living and would be game to be trained for a job working with driverless trucks. But in the meantime he argues that his current job is worth defending. “Being a driver is what I know,” he says. “That’s what I like about it.”
There are three schools of thought about the long-term effects of AI on employment. The first argues that advances in robotics will lead to improvements in productivity similar to those that occurred after other inventions—such as sewing machines, combine tractors, and washing machines, which freed up workers to do less repetitive (and better-paid) labor. The second school worries that the same technologies will require so few jobs that they’ll create a permanent underclass. The third school argues that it’s all hype and the advances are decades away.
Most people in Silicon Valley subscribe to either the first or second school. Much of the rest of the country, including many truckers, favor the third. “I can tell the difference between a dead porcupine and a dead raccoon, and I know I can hit a raccoon, but if I hit a porcupine, I’m going to lose all the tires on the truck on that side,” says Tom George, a veteran driver who now trains other Teamsters for the union’s Washington-Idaho AGC Training Trust. “It will take a long time and a lot of software to program that competence into a computer.”
The raccoon-porcupine divide is one of many in which computers may not work particularly well. But that doesn’t mean a system couldn’t be designed that would allow trucks to drive themselves most of the time. Viscelli, the University of Pennsylvania expert, says self-driving trucks will hit the road “in a matter of single-digit years,” and believes that they’ll allow the industry to eventually shed a few hundred thousand jobs.
Seltz-Axmacher acknowledges that companies such as his could ultimately make traditional trucking jobs a thing of the past, and he’s not sure what he or anyone else should do about it, beyond trying to be decent to the workers he employs now. He’s been reading about universal basic income—the idea, popular in techie circles, of simply paying everyone enough to live on.
But ultimately, Seltz-Axmacher believes, the tools he’s developing will be good for truckers. He cites a new book by Garry Kasparov, Deep Thinking, in which the chess great observes that middling chess players who play with the help of a standard computer are reliably better than either grandmasters or supercomputers by themselves. “I think humans and technology working together are always going to be better than either one alone,” Seltz-Axmacher says. “But maybe that’s just because I like humans.”