In the face of daily pandemic-induced upheavals, the notion of “business as usual” can often seem a quaint and distant notion to today’s workforce. But even before we all got stuck in never-ending Zoom meetings, the logistics and transportation sectors (like much of America’s economy) were already subtly shifting in the face of continuing advances in robotics, machine learning and autonomous navigation technologies.
In their new book, The Work of the Future: Building Better Jobs in an Age of Intelligent Machines, an interdisciplinary team of MIT researchers (leveraging insights gleaned from MIT’s multi-year Task Force on the Work of the Future) exam the disconnect between improvements in technology and the benefits derived by workers from those advancements. It’s not that America is rife with “low-skill workers” as New York’s new mayor seems to believe, but rather that the nation is saturated with low-wage, low-quality positions — positions which are excluded from the ever-increasing perks and paychecks enjoyed by knowledge workers. The excerpt below examines the impact vehicular automation will have on rank and file employees, rather than the Musks of the world.
Excerpted from The Work of the Future: Building Better Jobs in an Age of Intelligent Machines by David Autor, David A. Mindell and Elisabeth B. Reynolds. Reprinted with permission from the MIT PRESS. Copyright 2022.
THE ROBOTS YOU CAN SEE: DRIVERLESS CARS, WAREHOUSING AND DISTRIBUTION, AND MANUFACTURING
Few sectors better illustrate the promises and fears of robotics than autonomous cars and trucks. Autonomous vehicles (AVs) are essentially highspeed wheeled industrial robots powered by cutting-edge technologies of perception, machine learning, decision-making, regulation, and user interfaces. Their cultural and symbolic resonance has brought AVs to the forefront of excited press coverage about new technology and has sparked large investments of capital, making a potentially “driverless” future a focal point for hopes and fears of a new era of automation.
The ability to transport goods and people across the landscape under computer control embodies a dream of twenty-first-century technology, and also the potential for massive social change and displacement. In a driverless future, accidents and fatalities could drop significantly. The time that people waste stuck in traffic could be recovered for work or leisure. Urban landscapes might change, requiring less parking and improving safety and efficiency for all. New models for the distribution of goods and services promise a world where people and objects move effortlessly through the physical world, much as bits move effortlessly through the internet.
As recently as a decade ago, it was common to dismiss the notion of driverless cars coming to roads in any form. Federally supported university research in robotics and autonomy had evolved for two generations and had just begun to yield advances in military robotics. Yet today, virtually every carmaker in the world, plus many startups, have engaged to redefine mobility. The implications for job disruption are massive. The auto industry itself accounts for just over 5 percent of all private sector jobs, according to one estimate. Millions more work as drivers and in the web of companies that service and maintain these vehicles.
Task Force members John J. Leonard and David A. Mindell have both participated in the development of these technologies and, with graduate student Erik L. Stayton, have studied their implications. Their research suggests that the grand visions of automation in mobility will not be fully realized in the space of a few years.15 The variability and complexity of real-world driving conditions require the ability to adapt to unexpected situations that current technologies have not yet mastered. The recent tragedies and scandals surrounding the death of 346 people in two Boeing 737 MAX crashes stemming from flawed software and the accidents involving self-driving car-testing programs on public roads have increased public and regulatory scrutiny, adding caution about how quickly these technologies will be widely dispersed. The software in driverless cars remains more complex and less deterministic than that in airliners; we still lack technology and techniques to certify it as safe. Some even argue that solving for generalized autonomous driving is tantamount to solving for AGI.
Analysis of the best available data suggests that the reshaping of mobility around autonomy will take more than a decade and will proceed in phases, beginning with systems limited to specific geographies such as urban or campus shuttles (such as the recent product announcement from Zoox, an American AV company). Trucking and delivery are also likely use cases for early adoption, and several leading developers are focusing on these applications both in a fully autonomous mode and as augmented, “convoy” systems led by human drivers. In late 2020, in a telling shift for the industry from “robotaxis” to logistics, Uber sold its driverless car unit, having spent billions of dollars with few results. The unit was bought by Amazon-backed Aurora to focus the technology on trucking. More automated systems will eventually spread as technological barriers are overcome, but current fears about a rapid elimination of driving jobs are not supported.
AVs, whether cars, trucks, or buses, combine the industrial heritage of Detroit and the millennial optimism and disruption of Silicon Valley with a DARPA-inspired military vision of unmanned weapons. Truck drivers, bus drivers, taxi drivers, auto mechanics, and insurance adjusters are but a few of the workers expected to be displaced or complemented. This transformation will come in conjunction with a shift toward full electric technology, which would also eliminate some jobs while creating others. Electric cars require fewer parts than conventional cars, for instance, and the shift to electric vehicles will reduce work supplying motors, transmissions, fuel injection systems, pollution control systems, and the like. This change too will create new demands, such as for large scale battery production (that said, the power-hungry sensors and computing of AVs will at least partially offset the efficiency gains of electric cars). AVs may well emerge as part of an evolving mobility ecosystem as a variety of innovations, including connected cars, new mobility business models, and innovations in urban transit, converge to reshape how we move people and goods from place to place.
TRANSPORTATION JOBS IN A DRIVERLESS WORLD
The narrative on AVs suggests the replacement of human drivers by AI-based software systems, themselves created by a few PhD computer scientists in a lab. This is, however, a simplistic reading of the technological transition currently under way, as MIT researchers discovered through their work in Detroit. It is true that AV development organizations tend to have a higher share of workers with advanced degrees compared to the traditional auto industry. Even so, implementation of AV systems requires efforts at all levels, from automation supervision by safety drivers to remote managing and dispatching to customer service and maintenance roles on the ground.
Take, for instance, a current job description for “site supervisor” at a major AV developer. The job responsibilities entail overseeing a team of safety drivers focused in particular on customer satisfaction and reporting feedback on mechanical and vehicle-related issues. The job offers a mid-range salary with benefits, does not require a two- or four-year degree, but does require at least one year of leadership experience and communication skills. Similarly, despite the highly sophisticated machine learning and computer vision algorithms, AV systems rely on technicians routinely calibrating and cleaning various sensors both on the vehicle and in the built environment. The job description for field autonomy technician to maintain AV systems provides a mid-range salary, does not require a four-year degree, and generally requires only background knowledge of vehicle repair and electronics. Some responsibilities are necessary for implementation — including inventorying and budgeting repair parts and hands-on physical work—but not engineering.
The scaling up of AV systems, when it happens, will create many more such jobs, and others devoted to ensuring safety and reliability. Simultaneously, an AV future will require explicit strategies to enable workers displaced from traditional driving roles to transition to secure employment.
A rapid emergence of AVs would be highly disruptive for workers since the US has more than three million commercial vehicle drivers. These drivers are often people with high school or lower education or immigrants with language barriers. Leonard, Mindell, and Stayton conclude that a slower adoption timeline will ease the impact on workers, enabling current drivers to retire and younger workers to get trained to fill newly created roles, such as monitoring mobile fleets. Again, realistic adoption timelines provide opportunities for shaping technology, adoption, and policy. A 2018 report by Task Force Research Advisory Board member Susan Helper and colleagues discusses a range of plausible scenarios and found the employment impact of AVs to be proportional to the time to widespread adoption. Immediate, sudden automation of the fleet would, of course, put millions out of work, whereas a thirty-year adoption timeline could be accommodated by retirements and generational change.
Meanwhile, car-and-truck makers already make vehicles that augment rather than replace drivers. These products include high-powered cruise control and warning systems frequently found on vehicles sold today. At some level, replacement-type driverless cars will be competing with augmentation-type computer-assisted human drivers. In aviation, this competition went on for decades before unmanned aircraft found their niches, while human-piloted aircraft became highly augmented by automation. When they did arrive, unmanned aircraft such as the US Air Force’s Predator and Reaper vehicles required many more people to operate than traditional aircraft and offered completely novel capabilities, such as persistent, twenty-four-hour surveillance.
Based on the current state of knowledge, we estimate a slow shift toward systems that require no driver, even in trucking, one of the easier use cases, with limited use by 2030. Overall shifts in other modes, including passenger cars, are likely to be no faster.
Even when it’s achieved, a future of AVs will not be jobless. New business models, potentially entirely new industrial sectors, will be spurred by the technology. New roles and specialties will appear in expert, technical fields of engineering of AV systems and vehicle information technologies. Automation supervision or safety driver roles will be critical for levels of automation that will come before fully automated driving. Remote management or dispatcher, roles will bring drivers into control rooms and require new skills of interacting with automation. New customer service, field support technician, and maintenance roles will also appear. Perhaps most important, creative use of the technology will enable new businesses and services that are difficult to imagine today. When passenger cars displaced equestrian travel and the myriad occupations that supported it in the 1920s, the roadside motel and fast-food industries rose up to serve the “motoring public.” How will changes in mobility, for example, enable and shape changes in distribution and consumption?
Equally important are the implications of new technologies for how people get to work. As with other new technologies, introducing expensive new autonomous cars into existing mobility ecosystems will just perpetuate existing inequalities of access and opportunity if institutions that support workers don’t evolve as well. In a sweeping study of work, inequality, and transit in the Detroit region, Task Force researchers noted that most workers building Model T and Model A Fords on the early assembly lines traveled to work on streetcars, using Detroit’s then highly developed system. In the century since, particularly in Detroit, but also in cities all across the country, public transit has been an essential service for many workers, but it has also been an instrument facilitating institutional racism, urban flight to job-rich suburbs, and inequality. Public discourse and political decisions favoring highway construction often denigrated and undermined mass transit, with racial undertones. As a result, Black people and other minorities are much more likely to lack access to personal vehicles.
“Technology alone cannot remedy the mobility constraints” that workers face, the study concludes, “and will perpetuate existing inequities absent institutional change.” As with other technologies, deploying new technologies in old systems of transportation will exacerbate their inequalities by “shifting attention toward what is new and away from what is useful, practical, and needed.” Innovating in institutions is as important as innovating in machines; recent decades have seen encouraging pilot programs, but more must be done to scale those pilots to broader use and ensure accountability to the communities they intend to serve. “Transportation offers a unique site of political possibility.”
