photo by Flickr user zombieite
Self-driving cars are the future. Probably. Just maybe not as near a future as the automotive industry – and optimistic investors – had imagined.
A few years ago, the rise of driverless cars seemed not only inevitable but imminent. Google CEO Sergey Brin once predicted that “ordinary people” would be using fully autonomous vehicles by 2017. Uber CEO Travis Kalanick promised in 2016 that his company would get rid of human drivers by 2020. Various automakers heralded a huge shift in the type of vehicles they’d be making within a handful of years. Just a few years ago, the federal Transportation Department worried that technology would outpace regulation.
But more recently, a future without drivers seems much further off. In 2018, a self-driving car owned by Uber killed a pedestrian in Tempe, Arizona. That accident put a chill on not only Uber’s testing, but the development of other autonomous vehicles, as popular sentiment shifted toward greater wariness. Companies abruptly seemed to realize the high stakes that testing in populated areas might involve.
Popular distrust may also hinder the widespread adoption of driverless cars, even as the technology improves. Beyond worries about accidental collisions, security is a major concern. No passenger wants to be in a car when the driver is subject to hacking. AAA found that 71% of Americans surveyed were afraid to ride in a fully autonomous car. Industry experts suggest that this fear will decrease as people have more experience with these vehicles. But when and how most people will get that experience isn’t clear.
Beyond safety concerns, there are other practical considerations. As Timothy B. Lee observed for Ars Technica earlier this year, there are two phases of development for self-driving cars. The first involves making sure the vehicle obeys traffic laws, preserves the safety of passengers and pedestrians, and generally avoids obstacles. Some companies, including Waymo, have made it this far.
Yet the second stage may be equally important for widespread adoption. In stage two, vehicles will have to learn to make quick, accurate decisions in order to move as a skilled human driver might. Otherwise, the vehicle could have trouble navigating roundabouts, merging onto highways, or completing other potentially dangerous but essential elements of driving. A car that doesn’t master stage two may get you safely from point A to point B, but it would be a long and frustrating ride. As Sam Anthony, the chief technology officer of Perceptive Automata, told The Huffington Post, “The difference between a good self-driving car and a perfect self-driving car is massive.”
Creating a smooth and straightforward ride has proven more challenging than automakers initially anticipated. While driving on well-paved roads in good conditions is one thing, autonomous vehicles have had trouble with heavy rain and other challenging weather. Construction and badly maintained roads also cause problems. Self-driving cars will need to anticipate the actions of human drivers, cyclists and pedestrians, which may be hard to predict through programming. All of this means development has taken more time than companies originally projected.
Regulation is also a concern. No federal rules currently govern the testing or development of autonomous vehicles, though some states make it harder or easier for developers to operate. Carmakers also may try to influence lawmakers. Some have suggested that updating road infrastructure could mitigate some of the vehicles’ technological shortcomings. An unnamed automotive official suggested to The New York Times that Manhattan could install anti-jaywalking gates at each intersection. (I’ll let you imagine how New Yorkers are likely to take this idea.) How the interplay between government and business evolves may hinder or help the rollout of self-driving cars and trucks.
While boosters remain – Tesla CEO Elon Musk remains confident that driverless vehicles will shuttle people to their destinations by late 2020 – most industry insiders agree we’re decades away from fleets of cars that don’t need monitoring from a human driver. General Motors’ subsidiary Cruise said in 2017 that it would launch automated taxis in San Francisco by the end of 2019. This July, Cruise announced that it would not meet the deadline and did not set a new one. Waymo launched a taxi-like service in Arizona last year as promised, except that human drivers remain behind the wheel.
In light of all these setbacks, it seems reasonable to wonder: Just how far away are self-driving cars from commonplace use? What if they never arrive at all, at least to the extent developers have promised?
Self-driving cars are not an all-or-nothing proposition. Many vehicles on the road now boast “advanced driver assistance systems” that automate at least some tasks, such as adaptive cruise control or the ability to park with minimal driver input. Much like automatic transmissions and power steering, such features are likely to become an everyday part of how we drive as older cars leave the roads. Even if we don’t reach a point where there is no driver, the driver may be able to rely with increasing confidence on the car’s computing power.
SAE International, an organization that develops engineering standards in the automotive industry and others, describes vehicle automation on a scale of zero to five. At level zero, the human driver is responsible for all driving tasks. Most cars with “advanced driver assistance systems” would be around a level two, though level three automated vehicles may become common in the future. Think of the way autopilot systems today assist in most, if not all, aspects of flying a commercial jet. Level five would be a fully autonomous experience, in which the only thing a human needs to do is enter the destination. That has yet to arrive.
In some cases, full level-five autonomy may never be the goal. For example, most startups focused on long-haul trucking intend to leave urban driving to humans for the foreseeable future, even when trucks are ready to handle highways unaided. Yet for those invested in a completely autonomous future for cars, levels one through four likely aren’t good enough.
Based on indications today, I would not be surprised to see autonomous vehicles in more limited use in the near future. On a closed course under controlled conditions, they can work. For instance, Optimus Ride recently launched a shuttle service in the Brooklyn Navy Yard, which will carry passengers between the office park’s entrance and its new ferry dock. (For now, this service still involves human operators as a safety precaution.) Autonomous vehicles may be useful in areas that are otherwise closed to motor vehicles – say, moving people around amusement parks, airports or college campuses.
Investors, however, have plenty of reasons to be wary of any champagne-tinted predictions that traditional, human-driven cars will vanish. Self-driving cars are unlikely to revolutionize the auto industry in the next few years. They are not going to resolve the economics of ride-hailing services like Uber in the next year or two, either. In fact, like ride-hailing – as well as food delivery services and a variety of other sectors – self-driving cars are attracting a lot of hype but haven’t yet proven a profitable and reliable part of our future.
When it comes to truly driverless cars, the future may or may not be coming, but it certainly isn’t now.