There is something truly remarkable about autonomous driving technology. The thought that society may very soon experience a disruption in a basic function of everyday life—that is, driving with our own hands—is at first quite unsettling for many. But the promise of such an evolution can also stir up a bit of excitement. With the race towards commercialization well underway, President Obama remarked on the need to foster this nascent technology in a September op-ed for the Pittsburgh Post-Gazette. Delving into the promise that self-driving cars offer, he explains how they “have gone from sci-fi fantasy to an emerging reality with the potential to transform the way we live.” His outlook is shared by many, from optimistic politicians and ambitious industry players to eager consumers and technological experts.
Autonomous vehicles are more than simply the next revolution in the way we drive. They will invariably disrupt our patterns of life, opening the door to increased safety, improved productivity, and new logistical methods. Research from Morgan Stanley indicates that the autonomous industry may contribute as much as $1.3 trillion in annual savings for the nation’s economy. Displacing any need for a human driver, these vehicles rely on a complex system of various sensors attached to the car to visualize the surrounding area. Algorithms programmed for interactions with other vehicles, pedestrians, and objects then provide the car with a sense of how to navigate the road. The more diverse road conditions this kind of software encounters, the more capable it is of anticipating its surroundings. This is referred to as machine learning—there is no need to continuously program the software in order for it to expand its capabilities because it learns on its own. Possibilities to harness such innovative technology seem endless, as vehicle performance, driver behavior, and environmental details are all relatively straightforward applications of machine learning.
Some of the nation’s most inventive companies are already capitalizing on the rapidly progressing industry of self-driving vehicles. “A lot of different companies that wouldn’t have necessarily gone into the automotive space are jumping in with two feet. You’re getting talent coming from adjacent areas: robotics, artificial intelligence, and just programmers, because so much of it is software,” says Kathy Winter, the newly appointed vice president and general manager of Intel Corporation’s Automated Driving Solutions division. Intel develops some of the fundamental components for computational power in these vehicles, and teamed up with BMW and Israel-based Mobileye to complete a formidable alliance in the industry.
The trio will be racing to market with a diverse group of industry players. Beyond Intel, Alphabet Inc.’s Google and Elon Musk’s Tesla are two of the most prominent competitors, all of whom are seeking release dates in the near future. Some of the more aggressive projects are targeting a 2018 date, but 2021 is the most frequently cited year. Google and Tesla exemplify two distinct paths towards driverless vehicle technology and are among those spearheading the campaign for full commercial release.
For over seven years, Google has sought to develop fully autonomous technology from the ground up. As of today their test fleet stands at nearly three dozen prototypes on the road in several different locations throughout the country. Because of its other ventures, like Google Maps, the company has forged an initial advantage for itself in the industry. Its extensive data portfolio—including, for instance, Street View layouts—enables the self-driving project to draw on a vast amount of existing information. Google’s prototype is able to map out its surrounding area in great detail, determining and responding to even the most unforeseeable road events. Google has not yet publicly targeted a commercial release date for its fully autonomous vehicle.
Tesla’s AutoPilot feature, a recent addition to the Model S, is an example of the company’s incremental approach, which contrasts with Google’s desire to have fully autonomous vehicles before releasing anything. AutoPilot represents Tesla’s first attempt at commercializing semi-autonomous driving technology. It is essentially a glorified cruise control, barely reaching a level three autonomous classification by industry standards. This means that AutoPilot’s capabilities are not only limited to basic highway traffic situations, but also far below Google’s goal: full autonomy, which is level five on the industry scale. AutoPilot allows the Model S to steer, change lanes, and adjust its speed independently while traveling on the highway. Industry-leading software updates constantly improve AutoPilot’s capabilities, as Tesla is still in the process of beta-testing the technology.
The two projects differ in this fundamental principle guiding their competing projects. Tesla’s aggressive pursuit of semi-autonomous implementation led, this past summer, to the first fatality while using driverless technology. The incident serves as a stark indication of some unnerving questions central to this new technology, which had previously received little scrutiny.
At MIT’s AgeLab, researcher Bryan Reimer’s work on the dynamics of human behavior highlights the complex, seemingly paradoxical relationship between driver autonomy, driving performance, and distraction. Though distracted driving is a clearly hazardous endeavor, he maintains caution in merging the abilities of human driver and machine. First and foremost, distracted driving has catalyzed the troubling increase in traffic fatalities over the past year. Human error, according to the NHTSA, accounts for over 90 percent of traffic fatalities. Younger drivers are especially responsible for the uptick in distracted driving incidents. Tesla, like every other player in the industry, views this as an unacceptable trend. So why wait to implement autonomous vehicle technology if it can save even one life?
However, Reimer is also careful to note the potential dangers of autonomous driving, which he believes could worsen driver performance. With intelligent software commanding basic driving operations, drivers are almost incentivized to seek distraction. Joshua Brown, the man behind the wheel of a Tesla during a fatal crash on AutoPilot, was watching a film on a portable DVD player when his vehicle rammed into a fifty foot semi-truck. In a study by the Virginia Tech Transportation Institute, researchers affirmed Reimer’s worries about the impact semi-autonomous driving modes can have on attention spans. They found that drivers take noticeably longer off-road glances while behind a semi-autonomous wheel. Tesla warns all drivers that, even though AutoPilot can sufficiently control the vehicle in some conditions, their supervision is required at all times.
Mishaps with this new and imperfect technology are to be expected. In the case of Joshua Brown, Tesla’s sensors failed to discern the difference between the white truck and a bright sky. Every self-driving vehicle project is therefore pursuing robust testing to improve on, and ultimately overcome, these early flaws. Google has clocked over two million miles in road testing; Tesla is not far behind in its semi-autonomous endeavor. Exposure to various road and traffic conditions is crucial for enabling machine learning to reach its full potential. The more miles these prototypes drive, the more adept they become at navigating the road on their own.
Proving reliability with full certainty necessitates a prohibitively lengthy course of test runs. A study conducted by the RAND corporation concluded that hundreds of millions of miles need to be driven by autonomous vehicles before their track records can be compared with statistical confidence to those of human drivers. For example, experts say data released by Tesla are misleading because its vehicles are currently programmed to only negotiate basic highway traffic, not the more complex situations that an average human driver might encounter. Elon Musk, Tesla’s CEO, insists that its semi-autonomous technology is safer than the typical human driver, but he compares the two driving modes on this falsifiable statistical premise. Researchers urge Tesla–and all other industry players–to release more complete data on the performance of their autonomous vehicles so that the limitations of their technology can be better understood before commercial release. Many have looked to policymakers to mandate such a process.
The National Highway Traffic Safety Administration (NHTSA) recently established guidelines for the fledgling industry, communicating that the federal government is aiming for maximum flexibility in its policy approach. The agency is working with industry experts to accomplish this goal. Intel’s Kathy Winter notes the significance of NHTSA’s initiative, and echoes counterparts in the industry in her hopes that NHTSA “continues this trend of working with the industry and all aspects of the industry…to solicit feedback and be willing to adjust and make changes if something doesn’t make sense. With the technology moving so quickly, it’s difficult to sit here and imagine what we might need nine months or two years from now.”
A key aspect of the NHTSA guidelines focuses on data sharing, a “radically new” policy that is “truly for learning and not for profit,” as Winter sees it. As the competing projects gather information on the performance and limitations of their self-driving vehicles, the agency hopes they can collaborate and extract as much as possible from it. “Everybody would agree: better, safer, sooner,” says Winter. Although NHTSA outlined data sharing among its guidelines, the agency has yet to clarify the specific regulatory details. This means that data sharing has not yet been implemented, a potential hindrance to progress for everyone in the industry.
Though it established national guidelines, NHTSA deferred much responsibility to states, letting them experiment with different legislative approaches. California was the first of ten states to enact legislation allowing some form of autonomous vehicles. As a breeding ground for this emerging technology, the state has paved the way for extensive testing, but stipulated that a human must be present at all times behind the wheel. Michigan passed more ambitious policy that does not require a human’s presence in a self-driving vehicle. Arizona also recently positioned itself among the states most amiable to autonomous testing in an effort to cultivate a business-friendly environment; Governor Doug Ducey issued an executive order to accommodate self-driving technology and encourage the growth of this industry in Arizona. Intel, General Motors, Ford, and Google have since established testing grounds in the state, taking full advantage of Ducey’s pro-business, pro-experimentation initiative.
Though policymakers have been given free reign in this arena, many are confounded by the dizzying array of policy implications that the autonomous driving industry presents. On one hand, they seek to foster this sort of innovation within their state by allowing extensive road-testing. This is a crucial step towards gauging the safety of driverless technology. However, not everyone is on board with this experimentation. Some are reluctant to let these vehicles on the road until there is better understanding of the implications, as issues of liability and pedestrian safety persist. In a survey conducted by AAA, 75 percent of surveyed drivers stated they would be afraid to let an autonomous vehicle drive them.
As liability and safety questions linger, some legislative efforts have turned out to be a bit more hostile than friendly states of California, Michigan, and Arizona. Chicago and New York City are among these; the Chicago City Council is considering banning autonomous vehicles altogether. Such a sway in regulatory responses has engendered some concern among those looking to establish more uniform standards. To policy skeptics, these varying regulatory efforts are merely patchworks of legislation that could slow progress down the road. Others see these efforts as an indication of federalism operating as it ought to.
Regardless, political progress isn’t easy. A combination of an undecided electorate and befuddled policymakers creates prime conditions for profit-seeking companies to influence legislative realities. The Self-Driving Coalition for Safer Streets was officially formed in April of 2016. It boasts an impressive alliance of Google, Uber, Lyft, Ford, and Volvo, all of which share the vision that granting self-driving vehicles full autonomy is preferable to taking a more incremental approach. The Coalition tapped David Strickland, former head of NHTSA, to lead the charge against burdensome regulation. He seeks to influence policymakers at the state and local levels and has lobbied his former agency in an attempt to “develop appropriate policies, guidelines, and regulations that support the safe and swift development of full self-driving vehicles.”
The lobbying efforts of Silicon Valley titans and legacy automakers in this emerging industry have surely urged the political process along. Yet as these companies continue to capitalize on such a favorable environment, some of the most pivotal questions are pushed down the road. Integrating self-driving technology into a world of conventional vehicles poses questions that policymakers are still grappling with. Those states that have warmly embraced public testing on their roads will serve as models for the rest of the nation, as more skeptical observers wait to see how these policy experiments play out. No matter how muddled their efforts, though, the fate of this new industry does not seem to rest in the hands of bureaucrats.
American consumers seem increasingly likely to step up to the role of decision-maker. Without a consensus on the current safety of autonomous technology, they will have to decide for themselves how comfortable they are allowing the new technology to displace conventional driving. Policymakers can only guide them towards this outcome by regulating the industry; experts can only encourage them by assuring that ubiquitous self-driving technology will be safer than human drivers. Until consumer trust starts to improve, the success of this new industry will remain unpredictable. However, industry players can take a step in the right direction by pursuing shared and transparent data collection–such action would indicate that autonomous innovators are serious about earning the public’s trust.
Despite today’s obstacles, some of the nation’s brightest companies continue to push forward to deliver this promising new technology. Exciting opportunities are on the horizon; autonomous vehicles may very well be the next frontier in which America’s definitive capitalist spirit truly radiates.
The image featured in this article is licensed under Creative Commons. The original image can be found here.