As our lives feel increasingly like science fiction, our anxiety about the rise of the robots takes a variety of shapes and forms. The idea that robots will supplant human jobs and usher in an era of mass ennui and/or unemployment is one that's been in pretty regular rotation in recent decades—a Time cover from 1980 posed the question with an illustration of a multi-tentacled, all-purpose automaton worker; last month, in The Atlantic, Derek Thompson envisioned a "world without work." But do not fear (too much). Many of the most at-risk jobs will still require some level of human oversight.
Until fairly recently, the twitchy notion that more intelligent machines correlated with higher unemployment rates was dismissed by most economists as a "Luddite fallacy"; the belief was that advanced machines only enhance the labor market, or allow it to evolve, rather than supplanting human workers directly. Yet in the last ten years we've watched algorithms write news stories and learn to detect fraud. Uber's forthcoming line of self-driving cars threatens to displace a fleet of contractors the company itself created not five years ago. Given technology's rapid rate of acceleration, it's hard to predict what effect automation will have on jobs in the US even a decade from now. To wit: a recent study commissioned by the Pew Research Center asked industry experts—from science writers to Google execs—whether AI would displace more jobs than it created by 2025 and found them split on the matter nearly fifty-fifty.
In part, the divisiveness of the issue stems from a genuine confusion about the sorts of labor that can only be performed by humans. In the past, the consensus was that uniquely mortal qualities like empathy and creativity were necessary for the vast number of jobs, but those claims carry less weight in the wake of, say, Japan's investment in cute little robots that help grandpa navigate the nursing home or studies that claim people feel less judged when they unload on therapy software.
Techno-utopians may point to the Industrial Revolution, which created factory jobs as it displaced agricultural work, but as a program manager at NASA recently noted, what's happening now isn't so linear. "Robotics and AI are different," he told Pew, "Due to their versatility and growing capabilities, not just a few economic sectors will be affected, but whole swaths will be."
And keeping in mind that this is global capitalism we're talking about, it's unlikely that the oft-cited predictions of John Maynard Keyes—who believed mechanization would simply allow everyone a little more time to chill while robots did all the heavy lifting—will be the outcome. So, if the recent predictions of Oxford researchers are correct and nearly half of American labor will be automated by 2033, who's job is next on the chopping block?
Since the race to build and deploy self-driving vehicles began among tech companies like Tesla and Google several years ago, competition—and speculation about the wide-ranging cultural shifts such a change to American car culture would require—has been fierce. Depending on who you ask, mass-market self-driving cars could be on the road in just a few years, by 2020. And despite the fact that the transportation agency has yet to administer definite regulations, UC Berkley has tested a bus that steers itself by following a magnetic strip embedded in the road, as Google tests prototypes of its tiny self-piloting automobiles throughout Silicon Valley.
Some subway lines in major cities, including New York and Paris, have been selectively automated for years. In most cases, employees are retained to keep an eye on the automatic systems. In the United States in particular, the real risk of automation falls on private agents—New York's ubiquitous yellow cabs; personal chauffeurs and livery drivers; the already-precarious class of Uber contractors, the latter of whom CEO Travis Kalanick committed to replacing the moment it becomes viable. "The reason Uber could be expensive," he said last year, "is because you're not just paying for the car—you're paying for the other dude in the car."
In Rise of the Robots, Martin Ford speculates that self-driving cars produced and maintained by private tech ventures will include a centralization that could displace thousands of employees. "In the world that Google envisions," he writes, "robotic cars will be concentrated into fleets. Maintenance, repair, insurance, and fueling would likewise be centralized. Untold thousands of small businesses, and the jobs associated with them, would evaporate. To get a sense of just how many jobs might be at risk, consider that, in Los Angeles alone, about 10,000 people work in car washes."
It would appear that there are any number of good reasons for getting truckers out of the cabs of 18-wheelers; the long-haul trucking industry is a punishing and, at times, sinister profession. Labor Department data regularly ranks heavy load transport as one of the deadliest jobs in America; suffering long hours in the driver's seat with little access to heathcare and nutritious food, truckers regularly pull quite machinelike 14-hour days. The detrimental effects of the profession—and perhaps the kind of people it draws in the first place—have created a highway culture full of brutality and violence. But all those Amazon orders too big for drones must be fulfilled somehow.
There are currently more than 3 million full-time truckers in the US, but it's beginning to look as if this is the most recent blue-collar industry to fall under scrutiny as navigation and imaging technologies supplant human drivers. Earlier this year, the Freightliner Inspiration, the first self-driving truck, rolled down a highway, though its creators claim it'll be a decade before it's available to consumers. While fleets of self-driving trucks would have no problem cruising down America's freeways, they would still require humans to guide them from a highway exit to a delivery point—so like many workers threatened by automation, truckers would face significantly decreased hours and a wildly different job description. As Henning Schulzrinne, a professor at Columbia University, has argued, many "jobs won't be replaced, but rather be down-skilled or bifurcated into a small number of high-skill, high-pay and a much larger number of low-skill, low-pay positions" when such technologies are finally deployed en masse. And, as one blogger recently noted, the ripple effect is huge when you consider the number of American businesses built on housing and feeding a human workforce—robot drivers don't require truck stops, motels or diners.
Earlier this year, the FAA approved Amazon's request to continue testing delivery drones; the tension between government regulators and tech companies (as well as, it should be noted, the unsung heroes of Wisconsin-based Lakemaid Beer Brewery) over the safety and viability of such an unprecedented and rapidly expanding technology has been fierce. But regulatory hurdles won't deter companies looking to trade delivery drivers for remote pilots forever—if Jeff Bezos has his way, you'll be able to get packages under five pounds (think: N64 controllers, yoga pants) delivered to your door in 30 minutes or less. Mary Cummings, a researcher at MIT, has said she thinks drones will "augment the delivery world" rather than supplant it, likely owing to technological limits: drones can't deliver that love seat, and may have difficulty navigating dense urban areas.
Not to be outdone by Amazon, UPS and FedEx are both experimenting with drones, and the latter reportedly hopes to have three or four pilots minding the hundreds-strong "FedEx fleet" by 2020. The company's ground delivery department currently employs over 70,000 people driving 45,000 trucks, a number that would likely shrink significantly. Which isn't to mention harder-to-quantify impacts once drones are integrated into local supply chains: pray for the job security of your local pizza delivery kid. Or, your drug dealer.
We humans are a prideful species; those in the so-called creative industry never tire of reminding us that certain kinds of work require a human touch. True, the most vulnerable jobs in the brave new robot economy are rote physical tasks that garner less than a living wage, but that doesn't mean smarter algorithms can't do smarter work. And though it might be an inconvenient truth, there are plenty in the creative less-than-creative jobs.
Case in point: last year, BuzzFeed automated its first (slightly garbled) listicle with content pulled from the secret-sharing app Whisper. In terms of letting bots do some of the more basic aggregation journalists used to dutifully perform, they're far from alone. The Los Angeles Times has algorithms generating earthquake reports based on seismographic data and the AP used robots to report, first, on corporate earnings and, later, on college baseball games. In both cases, the generative software is provided with raw data, which it then spits out in natural (if cliche-riddled) language. As tech writer Kevin Roose said, referring to Automated Insights, the start-up that converts big data patterns into legible narratives, many of the "stories that today's robots can write are, frankly, the kinds of stories that humans hate writing anyway."
It's conceivable that automated writing software could free up writers to do more meaty, gratifying work like lovingly-crafted profiles and long-term investigative reporting. But it's also a classic mistake to assume that advanced technologies will replace, rather than merely alter, the professions they touch. Word processing didn't make paper obsolete, but it did change the publishing workflow forever, allowing writers to swap out entire paragraphs at the tap of a key.
Recently, Google released a set of images created by the neural networks that process its image search feature. They were horrifying: synthetic, undulating composite canines. Like a lot of robotically produced art, they tell us more about the inner workings of learning algorithms than they threaten the work of human artists. The same could be said of Aaron, the high-profile robot "painter" whose neon landscapes and floral abstractions were the result of nearly 40 years of continuous reprogramming and machine learning, or of more recent attempts to teach a piece of software to draw, like the portrait-generating Painting Fool.
Today, robots are used largely as inventive tools (drone-assisted paintings) or facilitators of finely wrought fabrication processes (3D printing). The fine arts market is one of the most manipulated in the world; much of the money circulating through it relies on fetishizing the genius myths of hyper-saturated artists like Damien Hirst and Jeff Koons. Given those conditions, fine art won't absorb the death of the author easily—but for art handlers and museum tour guides, it might be a different story. Even the legion of studio assistants that form the backbone of many artists' practices could be conceivably replaced by robotic counterparts. When Sol LeWitt described the method by which he created his large-scale, colorful installations, drawing detailed plans and having them executed by a small army of laborers, he did so through the analogy of composing a symphony. It's the same metaphor often favored by technologists to explain the impact of mechanization on the creative industry, which they frame as a process of freeing up great minds to orchestrate swarms of precise manual workers.
In graphic design and marketing, automated image generation is having something of a moment: M&C Saatchi recently launched a set of smart posters that use AI to change the tone and visual appearance of an advertisement. David Cox, one of the project leads, told Hopes&Fears that algorithms are already having an impact on his industry, "and it's certainly going to play an increasingly significant role in its future. But this doesn't necessarily suggest that there will be less jobs: it's more that the quality and nature of the jobs available will change. Automation technology within this industry—and beyond—will have the power to take on the more menial, time-consuming parts of our jobs, leaving us humans space to use our brains and be creative."
For cynics and doomsayers predisposed to imagining a future in which humans are useful only insofar as they're great at pressing buttons, the music industry provides a neat dystopic blueprint. Predicting the future of musicianship is difficult precisely because so much of it has been automated already: beat-matching softwares have diluted the skills required to be a DJ; synthetic production has, in some cases, supplanted the need for skilled musicians. A recent SNL skit lampooned the problem with one popular form of EDM, positioning an Aviici-clone onstage with nothing but a single button reading "Bass."
Save for a handful of notable exceptions—see David Cope’s fantastic experimental algorithm "Emily Howell," a machine learning program that scans classical music and writes its own compositions based on the patterns it detects, or Hatsune Miku, the holographic Japanese pop star whose songs are written by fans and performed using a voice synthesizing program—most machines created to replace human musicians are little more than party tricks. The real effect on jobs has been in studio and production work, made infinitely cheaper and easier with softwares like Pro Tools and Ableton. And in an industry that's growing less and less profitable by the year, what manager could resist the lure of such cost-cutting measures?
Christopher Stiener, author of Automate This, sees more than just studio jobs being performed by software in the future. "I think it's inevitable, in all industries," he says. He uses the example of Music X-Ray, a company that developed a sophisticated algorithm to isolate individual elements of a pop song—melody, chord progression, tempo, etc.—and match them with hits from the past, essentially predicting which musical unknowns will top the charts in the future, thereby automating the record industry's A&R process. He clarifies, "In Silicon Valley and the tech world in general, it's the creators of software who end up with the most money, the middle of the employment market is hollowed out. It's no different in the creative industry."
Steiner is correct to note that the creators of software and algorithms themselves are less likely to see their jobs being automated out of existence. As Paul Ford, who recently penned a massive cultural history-slash-explainer on coding told Hopes&Fears, "right now there's just need for more programmers. It's totally possible that you’ll be able to do more by saying, like, 'Siri, make me an app,' but they've been promising that for decades." In truth, a lot of what gets automated in programming has been built into the process to save developers time, for instance, running automatic tests every time a tweak is made to an app in order to make sure that it hasn't broken something else. As one software engineer at a prominent arts startup said to Hopes&Fears, "if human labor is being supplanted, it's the labor of clicking my mouse several more times."
While we might be a long way off from seeing the backend of websites being assembled without significant human input, one forthcoming project, The Grid, is attempting to use machine learning algorithms to automate the web design process. Billed as a "personal AI web developer," the program responds to a user's directives, whether they want an online store or a personal portfolio, and builds a site using what it's been programmed to interpret as "good design": basically, the colors and shapes that humans like to look at. According to its founder and CEO Dan Tocchini, The Grid will automatically parse content (links, photos, etc.) and integrate it into the site's design sensibility. Like many projects that threaten humanity's sense of individual creativity, The Grid promises it isn't here to put designers out of busienss—just to make them more ambitious. "Our AI is dependent on designers training it; the designers are still the masters, the AI just scales their effort," Tocchini has said. Still, given the rise of user-friendly interfaces, it isn't inconceivable that the average internet user would prefer to skip the cost of a high-end web designer and go with a fancy AI instead.
If you become a video game designer when you grow up, you're probably doing something right—the competitive, creative, and well-compensated jobs are consistently ranked as some of the best in the country, with the industry growing nearly 30% in the last ten years. Granted, with the rise of mobile gaming, some kinds of design remain more prestigious and satisfying than others. A computer certainly couldn't have invented the famously dense and atmospheric Myst though Flappy Bird might be another story. But as with just about every other job, there are rote tasks within game design that are easy to teach a computer, and the ways in which humans and AI will lean on each other in the profession's future is far from clear.
Mark J. Nelson, a researcher who studies video game systems and has written on automation and its impact on the design process, says that in some areas, automation is already very much a part of video game design. The best example he gives is three-dimensional trees—"Nobody makes individual trees for video games anymore," he told Hopes&Fears. Instead, most people used a piece of software called, appropriately enough, SpeedTree, which either automatically generates a forest or allows a designer to select which areas of a landscape should feature foliage. In the future, says Nelson, game design might be affected the same way CAD changed architecture—"it didn't really automate it," he says, "but it did change how architects work."
"There are some promising systems," he goes on, "Such as generating variants of a specific kind of board game. It's quite possible designers will use more of this in the future, as a two-stage process: the designer intents the game in rough generality, and then lets a computer look a millions of slight variants of the game to pull out a specific one."
At first glance, the idea that information security experts could be replaced by algorithms feels far-fetched; since the '80s, the spectre of the intuitive hacker genius has been omnipresent in pop culture. But in many cases, software is already doing some of that work: companies like PayPal and many banks use programs that comb customer transactions for patterns that could indicate fraud. And now that Defense Advanced Projects Agency (DARPA) is betting on the notion that intelligent machines could do the work it's currently employing hackers to do, the securities industry may be about to change significantly.
In 2013, DARPA announced the "Cyber Grand Challenge," an event in which teams of programmers will compete to create software that finds the most bugs in a type of hacking competition typically referred to as "Capture the Flag." Though the final showdown won't happen until the summer 2016, preliminary rounds have shown many of the AIs to be incredibly productive—this April, teams were about to confirm and patch twenty-three out of twenty-four bugs in the piece of software they were given. Mike Walker, a program manager at DARPA, has said that as automated machines get better at finding weaknesses, they will be invaluable; after all, computers can perform infinitely more calculations than humans in a given moment, and are capable of defending against attackers 24/7. "I believe the world series of hacking will soon be won by a machine," Walker has said. Human security researchers will still be necessary to solve creative problems—and obviously, to program this software—but in the next decade, there could be far fewer of them.