Takeaway: Recent high-profile accidents involving autonomous and semiautonomous vehicles raise new questions about the viability of driverless cars.

MARKET WATCH


March 18 marked a grim milestone for the automotive industry: That night in Tempe, Arizona, a pedestrian was struck and killed by a fully autonomous Uber test vehicle, the first such fatality in the United States. Less than a week later, a Tesla Model X driver crashed and died on California’s Bayshore Freeway while using the car’s semiautonomous Autopilot feature. Automakers and tech companies have big plans for autonomous vehicles, with many saying that consumers will be able to hail a driverless taxi within the next couple of years. But cracks are beginning to emerge in this roadmap. It could be decades before consumers will be able to buy a completely road-ready autonomous vehicle.

The auto industry is in the headlights following the recent tragedies. Much of what we know about the Uber crash comes from a disturbing video released by local police that shows the victim, 49-year-old Elaine Herzberg, emerging from the darkness into the path of the oncoming vehicle. The Tesla crash appears to mirror the 2016 collision that claimed the life of Tesla driver Joshua Brown. The latest victim, 38-year-old Walter Huang, crashed into a concrete divider while on Autopilot. The system seemed to have trouble handling that particular stretch of road: On multiple occasions prior to the crash, Huang had complained to Tesla about his vehicle veering toward that exact same barrier.

But automakers and tech companies won’t be deterred. Plenty of invested capital is riding on the impending autonomous vehicle revolution.

General Motors in December announced plans to deploy a large-scale fleet of driverless taxis in big cities by 2019. GM is also readying a vehicle with no steering wheel or pedals. The other market leader, Waymo (an Alphabet subsidiary), recently launched the first-ever ridesharing service to operate without a human behind the wheel. Ford, meanwhile, aims to have a fully autonomous taxi service ready for consumers by 2021. And then there is the “me too” crowd, consisting of companies like Toyota, Honda, and Renault-Nissan that have made vague autonomous plans.

Have Autonomous Vehicles Hit a Roadblock? - chart2s

On the surface, it seems as if automakers are all-in on bringing autonomous vehicles to market. But here’s the industry’s dirty little secret: Few insiders believe that it’s going to happen anywhere near as quickly as the PR and IR departments would have you believe.

Attendees of Bloomberg New Energy Finance’s Future of Mobility Summit (a gathering of the industry’s movers and shakers) were asked via live poll to estimate when consumers will actually be able to buy fully autonomous vehicles. The answer was damning: Nearly 75 percent said the milestone won’t be reached before 2030. Tech realists like Duke University engineering professor Mary Cummings have long known this secret: Cummings says that we’re 15 to 20 years away from a vehicle that “operates by itself under all conditions, period.”

Why exactly is full autonomy a long way off? Three main reasons: the technology isn’t ready; the regulations aren’t ready; and the public isn’t ready.

The most obvious challenge is the technology. It’s more than just the recent crashes: In the past two years alone, autonomous vehicles have also been caught running red lights and veering dangerously close to pedestrians. And that’s not counting the dozens of times that these vehicles would have crashed had a human test operator not intervened.

Sure, thanks to machine learning, the technology is constantly improving. But progress becomes exponentially tougher the closer you get to perfection. By now, autonomous vehicles have mastered most of the “easy” scenarios like reading road signs, identifying other vehicles, and tracking lanes in good weather. The remaining work requires getting machines to make sense of situations in which rules disappear and higher-order human judgment takes over. We’re talking, for example, about encountering bad weather, driving in construction zones, and figuring out ambiguous lane markings or traffic signals. (See: “Driverless Cars: Unsafe at Any Speed?”)

Another barrier is a lack of uniform law and regulations. Currently, 33 states have either enacted or introduced legislation dealing with autonomous vehicles, with the particulars varying widely from state to state. Some states, like Arizona, have no formal regulations whatsoever—a huge plus for companies seeking a safe testing haven. A more fundamental challenge deals with our legal and insurance framework. Today, if you are injured by a human driver, you can try to sue them. But if you are injured by a vehicle with no driver at the wheel, who then is liable? The automaker? The manufacturer of whatever piece of software or hardware “caused” the accident? Or maybe nobody, if automakers succeed in getting consumers to sign contracts waiving their right to sue in the event of an accident.

Even in the (unlikely) event that autonomous vehicle companies clear the above hurdles, there is one more barrier left: the general public.

According to Pew Research, 54 percent of adults are either “somewhat” or “very” worried about the development of driverless vehicles, compared to just 40 percent who are at least somewhat enthusiastic. Additionally, 56 percent say they would not personally want to ride in a driverless vehicle, compared to 44 percent who say they would want to do so. A separate survey finds that most consumers (64 percent) are worried about even being on the road with driverless cars. These survey results echo previous findings suggesting that consumers are skeptical of autonomous vehicles.

Have Autonomous Vehicles Hit a Roadblock? - chart3s

Part of the skepticism stems from our inability to process seemingly random twists of fate that take lives. In a car accident caused by a human, there is usually a logical, even relatable, explanation—like texting and driving. In a car accident caused by an autonomous vehicle, there is no such understanding. We may never know what exactly went wrong in the Uber fatality. By their very nature, machine-learning algorithms cannot reveal their intentionality. America’s ingrained car culture is also working against autonomous vehicles. Boomers and early-wave Gen Xers have participated all their lives in America’s century-long love affair with automobiles. (See: “America’s Car Culture Shifts Gears”). To them, analog cars bestow a sense of freedom that would be lost with autonomous vehicles.

To be sure, America’s attachment to the risk, independence, and privacy of car culture has a strong generational dimension. Millennials have shown far less attachment to car culture than their parents by any measure. Yet just 19 percent of 18- to 29-year-olds say they’d prefer to drive a fully autonomous vehicle, barely above average. Millennials also are more likely to believe that we need a whole new crop of rules and regulations before driverless cars are unleashed onto the public.

Have Autonomous Vehicles Hit a Roadblock? - chart4s

In the near term, full autonomy can achieve profitability in certain niche conditions. It is already proving itself on farms and in mines. It may soon begin to take on long-haul trucking, which features hours on end of monotonous, routine highway driving. But commercial production for everyday use by individual drivers? That’s a still a very long way off.

TAKEAWAYS

  • Don’t buy into the autonomous vehicle timeline. Two recent high-profile crashes involving autonomous and semiautonomous vehicles have touched off concerns over the state of the technology. Nevertheless, automakers and tech companies remain undaunted in their goal to bring driverless cars to the public within the next few years. This isn’t going to happen. It likely will be decades before consumers will be able to buy an all-situation, fully autonomous vehicle. Why? The technology isn’t ready; the regulations aren’t ready; and the public isn’t ready. These vehicles will be relegated to use in niche scenarios like long-haul trucking for the foreseeable future.
  • Take “miles per disengagement” figures with a grain of salt. The industry’s favorite statistic is its miles per disengagement rate. These figures often look impressive on the surface: Waymo vehicles were able to drive more than 5,000 miles between human interventions in California last year. But the ones defining and recording these “disengagements” are the companies themselves—and they’re not exactly impartial observers. Google doesn’t report disengagements in which the machine could have handled the situation.  Nissan, likewise, doesn’t count disengagements during the start or end of a test run. And then there’s the fact that these companies are often testing only in optimal conditions to begin with.
  • Note that semiautonomous vehicle makers are in a particularly tough spot. High-tech semiautonomous systems like Autopilot breed a false sense of security behind the wheel. Automakers that have spent big money implementing these systems are now trying to get drivers to pay attention. Tesla and Mercedes each warn drivers to take the wheel after a prolonged period of inactivity. GM’s Super Cruise mode uses an infrared camera to track a driver’s gaze, ensuring that their eyes are checking the road periodically. If not, drivers receive several warnings before the system eventually slows the car to a stop. “Semiautonomous,” as it turns out, is a misnomer—either a driver’s full attention is needed, or it isn’t.
  • Realize that the delayed autonomous timeline may benefit automakers. Paradoxically, the overall industry will likely be helped by long delays in the readiness of full autonomy. After all, if autonomous vehicles take off on schedule, it could trigger a vertiginous decline in total new car sales: According to the consulting firm RethinkX, the number of cars on the road would drop from 247 million today to a mere 44 million by 2030—an 80 percent plunge. In this scenario, only one or two global automakers would survive, except where they are nationally protected, and perhaps none of the survivors would be a U.S. firm or a traditional automaker.