The 100-Year History of Self-Driving Cars

What the long history of the autonomous vehicle reveals about its fast-approaching future

Vintage illustration of a family of four playing a board game, while their futuristic electric car automatically drives itself, 1957. Image: GraphicaArtis/Getty Images

The first self-driving vehicles were ships. After centuries of wrestling with wind and waves, ancient sailors devised contraptions that harnessed these forces of nature to fill in for man. They were simple but ingenious solutions, like the sheet-to-tiller system, which is still used today.

To rig it, you simply take the jib sheet (the rope that controls the smaller sail up front) and run it around a pulley and back across the deck. Finish by tying the bitter end to the tiller (the stick that steers the boat). Now, when a gust hits and the boat starts to round up into the wind, the jib will pull the rope around the pulley and yank the tiller, steering the vessel back the opposite way.

Tricks like this helped clever mariners relieve the fatigue of long shifts at the helm during the Age of Sail. You can use it to crack open a cold one and enjoy the spray as your yacht plows through the whitecaps like a train on rails. And while tillers were repurposed to steer the first automobiles, this old technique didn’t make the leap from sea to land — though we can imagine some frightful, fruitless attempts to make it do so. By 1891, the introduction of the steering wheel, by Benz, put the matter to rest.

On land, self-steering actually got harder when machines replaced animals. Motorization was a vast improvement over draft animals’ muscle power, but the gain came at the expense of brain power. It had long been common for riders on horseback, and even cart drivers, to fall asleep at the reins. Their dutiful animals would simply keep following the road or stop dead in their tracks.

Cars and trucks, however, needed drivers to guide them second by second. Their soaring popularity, combined with the growing risks posed by their weight and speed, birthed a variety of experimental self-steering schemes. One 1925 demonstration of a remotely controlled vehicle in New York City offered a glimpse of driverless autos to come, simultaneously tantalizing and terrifying the public. Cruising down Broadway before thousands of onlookers, the optimistically named American Wonder drove “as if a phantom hand were at the wheel,” reported the New York Times.

In the 1920s, motor vehicles claimed tens of thousands of lives annually — a death rate 18 times higher than today. This new technology promised to render city streets safe once again. But those hopes were soon dashed when the futuristic vehicle’s operators lost control — first at Sixty-Second Street and again moments later at Columbus Circle — before finally crashing the would-be wonder into another vehicle.

Despite this early misstep, the auto industry continued to daydream about remote-controlled cars. At the 1939 World’s Fair, the Futurama exhibit by General Motors featured an enormous motorized diorama of an American city. Free-flowing highways plied by self-driving cars, trucks, and buses crisscrossed bustling districts of slender skyscrapers. There was even a “traffic control tower” where, the future city’s designers imagined, dispatchers would direct the movements of tens of thousands of vehicles by radio. By the 1950s, guide wires embedded in the road surface had replaced radio as the preferred technology for remote-controlled vehicles. Ironically, it was RCA, the Radio Corporation of America, that staged the first successful demonstration of this approach in the 1950s.

These early prototypes showed the technical feasibility of automated driving, but their high cost and the lackluster demand for such features meant that neither radio-controlled nor wire-guided cars caught on. The price tag for guided-vehicle highways was thought to be as high as $200,000 per lane-mile. If fully built out, this road upgrade might have added more than 40 percent to the cost of building the Interstate Highway System, already the largest public works project in American history. Meanwhile, despite the dangers and drudgery of long or late-night drives, automakers were still riding a wave of consumer excitement about driving. They focused on producing powerful new cars that were exhilarating to drive.

These early dreams imagined a self-driving future based on external guidance. But by the 1960s, the focus had shifted to harnessing the new technology of computers to design vehicles that could truly, independently drive themselves autonomously, without outside help. At Stanford University, for the first time anywhere, researchers built robots that used cameras to see and computers to navigate. In highly controlled experiments, these early droids followed white lines and avoided obstacles placed in their path.

Self-driving wasn’t confined to the laboratory for long. CPUs and image-processing techniques improved, so that by the late 1970s engineers at the University of Tsukuba’s Mechanical Engineering Lab were able to test the world’s first self-driving passenger vehicle, on Japanese roads. Traveling at speeds up to 20 miles per hour, these first AVs used two video cameras to visually detect street markings. In the 1980s the action moved to Europe, where Ernst Dickmanns, a professor at West Germany’s Armed Forces University, retrofitted a Mercedes-Benz van with self-driving gadgets of his own design, launching a decade-long collaboration with auto giant Daimler.

Finally, it was the Americans’ turn, as Carnegie Mellon University took the lead in the 1990s. As the competition to build self-driving machines spread worldwide, the software improved quickly and computers got ever faster, unlocking new possibilities. By the decade’s end, the first cross-country trips under automated control — in the U.S., Germany, and Japan — were in the record books.

The most intense period of AV development was still to come. In the early 2000s, the Pentagon took a growing interest in this emerging technology. To focus the efforts of scattered research groups and catalyze stronger ties with the defense and auto industries, the Defense Advanced Research Projects Agency — the U.S. military’s most independent research-funding arm — organized a series of open competitions in 2004, 2005, and 2007. These “Grand Challenges,” as they were called, offered millions of dollars in prize money and priceless prestige, and attracted dozens of teams from academia and industry.

Putting their best hardware and software to the test, the competitors watched from afar as their AVs tried to traverse both open country and more suburban settings on an abandoned military base. The 2004 race ended without a winner — none of the entrants reached the finish line. But a year later, Stanford University’s winning vehicle claimed the $2 million prize.

The DARPA contests accelerated the development of driverless vehicles. Stanford’s first-place finish in 2005 was the result of its pioneering use of machine learning, an A.I. programming technique, in processing road imagery. But more important, the contests focused attention on the emerging technology’s possibilities. No one was shocked by the military’s rising interest in AVs. But it was the potential civilian applications that set off a sudden wave of speculation. For the first time, the practical commercial use of self-driving technology seemed within reach.

It was a wake-up call for the auto industry. But not everyone heard it. Most companies were preoccupied with the financial crisis of 2007–2008 and the global recession that followed. U.S. automakers in particular were hamstrung when it came to capitalizing on the opportunity of AVs, which would require substantial further investment for the journey from lab to market. The automakers were going bankrupt or getting bailed out by the federal government. Instead, Silicon Valley moved forward. By 2009, the head of the winning Stanford University team, Sebastian Thrun, was leading a new self-driving-car project at Google. The search giant had bet big on Android, its highly successful operating system for mobile phones. Cars could become the next big computing platform, it seemed. Could Google stake a claim on the future of automotive software? It appeared to be a smart bet, bolstered by CEO and cofounder Larry Page’s lifelong interest in AVs.

A new Google self-driving car is on display at Google X in Mountain View, California on May 13, 2015. Photo: Kim Kulish/Getty Images

Google’s move took a few years to sink in, but once it did, all hell broke loose — not only in the car business, but in the computer and cab industries as well. Suddenly, every major automaker, every ride-hail company, and competing cloudware giants like Apple hastily mobilized efforts to develop self-driving vehicles, too. When in-house projects failed to produce convincing results, many companies simply acquired promising startups to get hold of the needed technology instead. In a two-year period during 2016 and 2017 alone, some $80 billion surged into self-driving vehicle technologies.

The biggest deal, Intel’s panicked 2017 acquisition of computer-vision pioneer Mobileye, an Israel-based maker of computer-vision systems, was valued at an eye-watering $15 billion. As this flurry of mergers and acquisitions unfolded, the web of partnerships and cross holdings linking automakers and the tech sector grew ever more tangled. Two of the world’s biggest consumer industries — computers and cars — had seen their future in each other. But they couldn’t decide whether they wanted to get together or gobble each other up.

By 2018 the hard work and high finance had paid off. In December, Google spin-off Waymo quietly unwrapped the world’s first truly self-driving taxi service, in Chandler, Arizona. More than 40 years after the first AV test-drive at Tsukuba, and nearly a decade after recruiting Thrun, the company started taking requests for driverless rides through the Phoenix suburbs. Reports said the tech giant had set aside more than $10 billion to build out its self-driving empire. At last, it seemed, the long and painful birthing of the AV was finally over.

“There is hardly a task that horse-drawn vehicles can do which cannot be done as well, and possibly better, with automobiles,” reported the New York Times on January 12, 1903, as one of the world’s first big auto shows opened its doors inside Madison Square Garden, then located at Twenty-Sixth Street and Madison Avenue. The Times was still at it a century later, this time hawking the engineering marvels of the self-driving age with a similar enthusiasm. “On my fourth day in a semi-driverless car,” wrote columnist David Leonhardt in 2018, “I was ready to make a leap into the future.”

The paper of record isn’t alone. Much like the automobile, AVs have unleashed bold speculation about the new technology’s benefits to individuals and society. But what does that future promise?

First, self-driving technology can eliminate nearly all of the deaths caused by automobiles, say its champions. An estimated 60 million people were killed in motor vehicle crashes in the 20th century. That’s more than all of the military and civilian deaths during World War II. But even as cars have become much safer, the killing continues, as motor vehicles spread to new countries where skilled drivers and traffic regulations are in short supply. As auto use booms in China and India, more than 1.4 million road deaths occur worldwide every single year — stealing enough souls to fill a city the size of Dallas, Texas; Birmingham, England; or Kobe, Japan. The vast majority of these crashes would have been prevented with self-driving technology, advocates claim.

Second, AV boosters boast, traffic congestion as we know it will disappear. The economic toll of overcrowded roads is enormous, and is easier to measure than ever, thanks to location-tracking devices embedded in ubiquitous mobile phones. Using the vast troves of travel records these phones leave behind, telematics firm Inrix estimated that in the U.S. alone, the cost of drivers’ time wasted in traffic was over $305 billion a year, or nearly $1,500 per driver. The argument for AVs is that software-piloted cars can safely pack more cars closer together at highway speeds, thanks to faster braking reflexes. But AVs might also reduce some bottlenecks by simply spreading human populations farther apart, splaying settlements out over a wider expanse of land. When passengers in AVs can use travel time for work or leisure instead of keeping eyes on the road, the thinking goes, longer rides to less-congested areas won’t be a bother.

Third, no one will be left behind by AVs, advocates hope. Cars expanded mobility for hundreds of millions of people in the 20th century, but when the automobile’s success dispersed the population and siphoned funds from mass transit, many found themselves facing new barriers to freely getting around. In the U.S. alone, more than 25 million people have disabilities that limit travel — nearly one-sixth of the workforce. Not only will AVs bring automobile travel to those physically unable to drive, it is believed, they will open up new travel options for the very old, the very young, and those who can’t afford cars of their own. As disabled people come off the sidelines and enter the workforce, as senior citizens get easier access to medical care, and as children enjoy access to a wider range of educational and enrichment opportunities, the social and economic benefits could be enormous.

When will this utopia arrive, you ask? Today AVs are still a novelty. Despite all the hassles, dangers, and drudgery of driving, we remain the most cost-effective “technology” suited to the task. By the time you read this, in the early 2020s, even if the wildest predictions come to pass, there will still be fewer than one million truly self-driving vehicles plying the world’s highways, streets, and sidewalks. But AVs’ numbers are destined to grow quickly as the decade rolls on. By 2030 the global headcount of smart cars, trucks, and buses could creep into the tens of millions. They’ll share the road with some two billion human-driven cars and trucks (give or take a few hundred million). Even then, it seems, AVs will be but a rounding error in the global population of automobiles. But the revolution will strike with surprise, surgical precision, and overwhelming force. As cyberpunk novelist William Gibson once famously said, “The future is already here — it’s just not very evenly distributed.”

Pilot models of the Uber self-driving car are displayed at the Uber Advanced Technologies Center on September 13, 2016 in Pittsburgh, Pennsylvania. Photo: AFP/Stringer/Getty Images

The first changes we notice will occur in taxis. Most market analysts agree that all taxis in the industrialized nations will be automated by 2030. In the U.S., that’s 300,000 vehicles. Add in all the Ubers and Lyfts and the total is closer to 1,000,000 in all. Swarming from our airports and resorts through our most beloved downtowns, driverless cabs could become the face of automation for a generation, and the gateway drug to driverless mobility for billions of passengers every year. The arrival of driverless cabs could radically change consumers’ perception of cars. When computerized chauffeurs are a tap and a swipe away, and robotaxi rides are dirt cheap, people may opt out of auto ownership altogether. If we make the shift en masse, far fewer vehicles will be needed to move the same number of people that private cars do today.

But this silver lining may not come to be. Automation will also make private automobiles more useful, and software will radically reduce the hassles of ownership. Think about it for a moment. Automated cars will do more than drive for you — they’ll also park themselves, take themselves to the garage for fuel and repairs, and pay their own insurance bills (with your money, of course). It’s entirely likely that we’ll simply swap our stupid cars for smart ones, and go on cruising around as we have.

In the long run we’ll likely see a mix of both worlds. By 2040, even if shared AVs take over and new-car sales fall by 50 percent — a sea change, indeed — automakers will still be churning out some 30 million self-driving cars worldwide every year. Half will end up in China, another quarter in America, and the rest scattered across the EU, Japan, and emerging markets. Yet even as the business of making cars shrinks, the business of using cars — and vans, and scooters, and everything else that goes — will grow. What’s left of today’s $2 trillion global auto manufacturing industry will be subsumed into a much larger market for “personal transportation services” that’s projected to reach $7 to $10 trillion a year by mid-century, roughly the size of the entire EU economy today. Waymo alone wants to capture a $1.7 trillion annual share by 2030. But Uber, Amazon, and Alibaba — not to mention Ford, GM, and VW, among others — aren’t ceding this new frontier without a fight. They have their own designs on the service businesses of the self-driving future, too.

So while the driverless revolution starts with a trickle, before long that slow drip will become a torrent. By 2050 or thereabouts, most human-driven cars will be gone. A smaller, smarter fleet of self-driving vehicles of many shapes and sizes will have replaced them. Some will be private, some will be shared. Some will move a single person, some will haul a hundred or more. Many won’t carry anyone at all, and instead will busy themselves with shuttling around an unceasing flood of goods unleashed by the triumph of online shopping. Some will help us by simply watching over our urban world or directing traffic. All told, our diverse fleet of AVs will log vastly more miles than our cars do today.

It’s tempting to see the driverless revolution as a repeat of our 20th-century experience with cars, only on a larger, computer-choreographed scale. But nothing in our past can prepare us for what lies ahead. At full tilt, the pace of change will bewilder us. In the U.S., full motorization took about 60 years — from roughly 1920, when cars started arriving in cities in large numbers, to 1980, when metro areas everywhere started to choke on their vast numbers. The next 40 years, from 1980 to 2020, was a period of saturation.

The average number of hours spent in traffic by commuters nearly tripled, and the economic cost of traffic congestion grew tenfold, to $166 billion annually. We have spent much of this time seeking ways to curb auto use and invest in alternatives.

But automation could play out in as little as 20 to 30 years — the span of a single generation. If our history with the automobile does teach us anything — it is that the future we find in the driverless revolution won’t be the one we expected.

From Ghost Road: Beyond the Driverless Car published by W. W. Norton & Company. © 2020 by Anthony M. Townsend

Forecaster and urban planner —

Sign up for Pattern Matching

By OneZero

A newsletter that puts the week's most compelling tech stories in context, by OneZero senior writer Will Oremus. Take a look.

By signing up, you will create a Medium account if you don’t already have one. Review our Privacy Policy for more information about our privacy practices.

Check your inbox
Medium sent you an email at to complete your subscription.

The undercurrents of the future. A publication from Medium about technology and people.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store