Will automation kill as many jobs as is feared? A widely cited Oxford University study predicts that 47% of jobs could be automated in the next decade of two. Price Waterhouse pegs the U.S. risk at 38%. McKinsey estimates that 45% of what people are paid for could be automated using existing technology!
No less than Tesla's Elon Musk, Bill Gates, and Stephen Hawking fear the loss of jobs will cause world cataclysm.
In the coming two decades, in what situations will humans still be preferable to an automated solution? Might that even offer implications for how we might live today?
Lower-level jobs at risk
Let’s start with jobs likely to be eliminated, starting with the present and with those lower-level jobs.
Already, don’t you prefer a ATM to a teller, self-checkout to the supermarket checker, drive-through tolls rather than stop for the toll-taker, automated airline check-in rather than waiting for a clerk, shopping on Amazon rather than fighting traffic, parking, and the check-out experience with a live clerk, assuming the store has what you want in your size? Indeed, malls are closing while online retailers led by Amazon are growing.
Further killing employment, the costs of hiring a person are increasing. Imagine you were running a business and deciding whether to hire more people to automate in the face of increased Social Security limits and Workers’ Compensation costs, paid family leave, ObamaCare or its replacement, living wage ordinances, OSHA requirements, and employee lawsuits. All that makes employers more likely to invest in automated solutions. After all, apart from the cost saving, robots and computers never sleep, take a day off, or come in late, Automated solutions are competent without training (and machine learning is making automation self-teaching,) make fewer errors, never quit and have to be replaced and retrained, never steal from the employer, never go on strike, never sue for sexual harassment, racial discrimination, or wrongful termination, and are never annoyed at a co-worker, customer, or you. Wouldn't you be automating as much as possible?
As minimum wage and mandated benefits rise, fast-food restaurants especially are accelerating use of, for example, order-taking kiosks, which McDonald's is rolling out in 2,500 stores, robotic burger flippers and fry cooks, even pizza, ramen and sushi makers. Even that fail-safe job, barista, is at-risk, Bosch now makes an automated barista. Mid-range restaurants such as Olive Garden, Outback, and Applebee are replacing waiters with tabletop tablets. Will you really miss having your conversation interrupted by a waiter hawking hors de oeuvres and expecting a 15+% tip? If you owned a fast-food franchise, mightn't you be looking to replace people with automated solutions? Can it really be long until there are completely automated fast-food and even mid-range restaurants? Whoops, it's already happening: Momentum Machines is opening 100% robot-run burger joints. That's also true at Eatsa, a healthy-fast-food chain. Sure, some patrons of high-end restaurant may want to interact with a human waiter: "Hi, my name is Sean and I'll be taking care of you this evening. Our specials tonight are..." But even in high-end restaurants, as tabletop-tablet ordering becomes more prevalent, most people will probably prefer to spend that time interacting with their dinner companions than with Sean.
Retail clerks also need beware. For example, Amazon now has cashierless, no checkout-line stores, Amazon Go, which it likely will install in its Whole Foods stores. Lowes is using LoweBots in some stores for real-time inventory control to ensure shelves are full and even to answer, in multiple languages, customer questions: Want the right nut for a certain bolt? Put it in the door and LoweBot will tell you the right item and walk you to it.
But what about clothing store clerks? With Amazon's StyleCheck, you use your phone's camera to take a photo of you in two outfits and artificial intelligence opines on which looks better on you. And now, Amazon Prime Wardrobe allows you to try on 15 items before actually buying. Pepper will not only answer your verbal questions, it will "read" your facial, emotional reactions and respond accordingly. It may even dance for you! A Palo Alto store that is using it found that it increased store traffic 70%. When the novelty fades, who knows? But by then, Pepper may have a whole bag of new tricks that a workforce of humans couldn't begin to duplicate. That vitiates one of brick-and-mortar stores' last advantages.
In short, robotics promises to kill many of the retail sales clerk jobs. From a humanistic perspective, one cannot oppose wanting all workers to make a living wage such as $15. But that would accelerate employers' decisions to hire a robot.
Of course, manufacturing is ever more automated. An MIT study finds that 670,000 manufacturing jobs alone have already been lost to automation. The number is expected to quadruple in the next decade alone! For example, a New York Times headline reads: “U.S. Textile Plants Return, With Floors Largely Empty of People.”
Robots are already being used as security guards. There are humanoid robots that can move heavy boxes, walk in uneven snow, and get up, not annoyed when thrown to the ground and who won't sue for failure to supervise or an OSHA violation. Then there's the Tesla plant filled with robots but few people—No more “Monday morning cars” built by workers who are hung over or who come to work under-the-influence. Companies have long-used robot-assisted warehouses but a recent Wall Street Journal report sees people-free warehouses coming. Someone who worked at a Ford plant told me that some assembly-line workers came in stoned and deliberately misassembled cars because they thought that was funny. The future of manufacturing? Well, all the major shoe companies are experimenting with 3D-printed custom shoes—Bye-bye offshoring but bye-bye jobs. And MIT researchers have created a robot that can 3-D print an entire house in hours! Construction workers will not find that funny.
Agriculture too is ever more automated—fewer stooping farm workers, more machines to till, plant, water, feed, and harvest. In Japan, the first fully automated lettuce farm will begin operations later this year.
Within a decade, we’ll likely have autonomous-driving trucks, buses, Ubers, and trains that are safer than human drivers. Autonomous truck drivers will never tailgate, cut you off or flip you off, drive recklessly under or not under the influence. They will never call in sick, subjecting everyone on the road to perhaps less trained substitute drivers.
Professions at risk
But what about professional-level employment? Well, millions of people who do their own taxes use TurboTax when, in the past, they needed an accountant. Even accountants use software to increase accuracy and speed. There’s little doubt that such software will grow ever more powerful, thereby further reducing the number of accountants needed. Professions that require much routine work are particularly vulnerable, and accounting is filled with that. Accounting jobs will remain mainly in strategic roles and for people to provide a "sanity check" when the computer spits out an anomalous number.
When I first bought shares of stock, I had to call a stock broker and pay $100 to $200 per trade for those screaming yellow-shirted floor traders to execute it. Now we buy stock online for $5 to $10 per trade, executed more accurately by a computer. Anyone want to bring back the old way? Bye-bye lots of jobs as stock broker and floor trader.
We’ll need some doctors, nurses, etc. but the speed and accuracy of computer-assisted diagnosis and treatment will reduce the number needed. For example, computers already diagnose cancer better than pathologists do. Robots are also reducing the number of surgeons and other medical professionals needed in the operating room.
Same is true of lawyers. Human nuance is required, whether in drafting a contract or appearing in court, but fewer attorneys will be needed. Already, increasing amounts of legal work—for example, sifting through massive databases to identify relevant cases, statutes, or contracts, which used to be done by teams of lawyers and paralegals—is done by computer. JP Morgan uses computers to analyze commercial loan transactions that used to require thousands of lawyer hours. A Deloitte study predicts that within two decades, 39 percent of legal jobs (including lawyers, paralegals, and legal assistants) are at risk of automation. CNBC has profiled a number of companies, including IBM, that are hard at work bringing artificial intelligence to the legal profession. A silver lining may be that automation's cost saving may make legal services more affordable.
Instead of hiring architects for tens of thousands of dollars, many people are opting to spend just a few hundred bucks to instantly download any of thousands of often award-winning house plans which, if needed, can be inexpensively customized to suit. Far fewer architects needed.
But what about those millions of other jobs that require human judgment? Yes, people will probably always be used to resolve disputed computer-based decisions, but that’s only a fraction of the people currently working in those fields. After all, The Golden State Warriors and Boston Red Sox used so-called MoneyBall, computer-based decision-making, including for hiring, to win world championships. Software developed at MIT was better than the tenure committee at predicting which young scholars would produce the most successful research. Robots are already being used to interview job applicants. (Could that mean that future hiring and promotion decisions will be more accurately made (and less subject to accusations of racism and sexism? if done at least in part by computer?) The Harvard Business Review reported that Fidelity and Vanguard are working on automating not only back office jobs but investment advising. BlackRock, the world's largest fund company has replaced seven of its 53 analysts with AI-driven stock-picking.
What about journalism? Even in major media outlets, many journalism jobs have already been lost to the armies of people willing to write for free. In addition, software such as Quill can replace some human journalists. It collects data, analyzes it, and writes a story in seconds, and the articles are more comprehensive and less biased than a human journalist's. The Associated Press already uses computer-generation for corporate earnings and minor-league baseball articles.
How about teachers? Perhaps no matter how immersive, how interactive, how individualized, how cute the artificial-intelligence-based "teacher," some students, especially kids, will want a live teacher to accompany a SuperTeacher humanoid. But automated teachers may produce, at lower cost, better results than obtained by the nation's live teachers, notorious for variation in quality. (Think, for example, about the teachers you've had for math.) Of course, automation, in the form of online courses, is already replacing many live teachers and trainers. For example, GE is using online courses to train hundreds of thousands of employees.
But what about the arts? There already is computer-generated classical music. There's artificial intelligence that can reproduce the style of many famous painters. Automated artists do have a limitation: At least until self-teaching computers advance enough, computers will only be able to reproduce versions of what human programmers program in. So creative jobs will remain for those rare birds who can create a cutting edge that’s not only different but superior to or at least more popular with customers than are computers’ creations. Alas, even that is at-risk. For example, already, IBM's Watson has reviewed thousands of songs to identify the sounds people like best and has reviewed thousands of major publications and websites to identify the themes that move people and thus the computer created a hit song, "Not Easy."
But what about romance? Is the premise of the movie, Her, permanently science fiction, in which a person prefers a computerized lover, not just because “s/he’s” great in bed but is a better listener and always eager to please? At a London university conference on sex robots, many people already said they'd prefer sex with a humanoid. Robotic lovers are also being touted for people having difficulty finding a partner, the disabled, and sex therapy patients. In additions, robots can be a remedy for the binary assumption—that we’re man or woman, homosexual or heterosexual—Robots can be pansexual.
There's even a virtual wife called GateBox already on the market in Japan, to address the famously long-working Japanese men with too little time and energy left to meet a real spouse. Most of the first customers have been men who felt too unattractive to meet a human spouse. For now, Gatebox is rated PG but I wouldn't bet on that lasting.
The remaining jobs
In such a world, how can a human justify asking to be paid to work?
In a column in The Guardian, Stephen Hawking wrote, "...artificial intelligence is likely to extend this job destruction deep into the middle classes, with only the most caring, creative or supervisory roles remaining."
More specifically, what sorts of jobs will remain for humans?
Well, certainly, we’ll need some brilliant people to develop technology, plus some hands-on types to maintain them. Among the most in-demand of these will be machine learning programmers who embed artificial-intelligence data security.
There will be jobs for connectors: people with the instincts to intuit what's not said and persuade, motivate, and negotiate. So the job market should remain good for so-talented business developers, managers, and salespeople of complex items, from industrial robots to major-gifts fundraisers. A word about the latter. Nonprofits’ lifeblood is money. And not surprisingly, among the best paying non-profit jobs is development specialist, people who specialize in extracting maximum dollars from wealthy individuals and corporations.
Higher education administrator. While campus-based education will likely shrink in favor of next generations of online courses, campuses may exist for a decade or two, especially serving undergraduate students, whose parents see college as a halfway house been the protection of childhood and the independence of adulthood. Campus jobs offer a stimulating environment and a relatively short work year. Don’t have a PhD? That’s often not required in, for example, student affairs: orientation, housing, and extracurriculars.
As mentioned, we’ll likely always need human judges to appeal or validate outlier computer decisions Also, program evaluators should be able to find work. Whether in the private or public sector, most innovative programs require an evaluation. We'll still need courtroom judges, although perhaps fewer. For example, there already is software that assesses a defendant's recidivism risk. It's not yet valid enough but it portends what's to come.
We'll need entrepreneurs to create simple, small businesses to meet local needs that can't be met by computer, for example, handypersons, tutors, and personal chefs.
We’ll need some fiction writers, mainly for the immersive interactive environments that most apartments will have: all four walls, floor, and ceiling can be screens. That said, there already is AI software, Angelina, that automatically develops video games. They're crude...for now.
We'll need superstar athletes and performers. Many people crave idols, and computer-generated sluggers, songs, nor animated characters may quite cut it. As today, most people will do creative and athletic activities for love but, alas, not for money.
China is already using robots for child and elder care, and in the U.S., Mabu is now being trialed on human patients, can read patient facial expressions, remind about medication, and notify the doctor on noting anomaly. Nevertheless, we'll need caretakers: companions, home-health aides, even dog sitters, to provide warmth that even an infinitely patient robot can't provide.
We'll likely need spiritual leaders. Traditional church attendance is down, perhaps permanently as fewer people believe in a God. But the need for community and a sense beyond the quotidian should result in continued need for spiritual leaders.
And yes, we'll need counselors and psychotherapists. While artificial-intelligence-based therapy is under development, the level of nuance in the excellent counselor will likely be refractory to computerization. Pick a niche likely to remain in-demand, for example, dating, parenting, menopause, or anger management. Or choose a niche likely to burgeon: interracial relationship counseling, transgender counseling, immigrant counseling.
And we'll need or at least have politicians. But given our democracy's lack of success in picking competent, ethical politicians, 20 or 30 years from now, is it absurd to think that an artificial Intelligence-driven humanoid candidate might run for office? I can see it now: R2-D3 versus Chelsea Clinton. Or can you imagine a human candidate who was gene-edited for high intelligence and ethics?
More broadly, it may be safest to select a career from one of these areas:
Government employment. Government has the deepest pockets and is relatively loose with its money because it can and does raise taxes or take on more debt. For example, the federal debt doubled in the past eight years to $20 trillion. So it's no surprise that pay is good for government employees. What may be surprising is that a just-released Congressional Budget Office study found that, for the same work, federal employees now are better paid than their private-sector equivalents. And the job security of government work is nonpareil plus, for political reasons, government is less likely to convert so many jobs to part-time/temp, to offshore or in-shore jobs to low-cost states. Of course, government isn't just federal. There's state, county, and city governments, plus a panoply of other planning and oversight entities. For example, in my locale, there are the Bay Area Transportation Management District, the Bay Air Quality Management District, and the East Bay Parks District, each with their own assemblage of employees.
Technology jobs. Per the above cited review of the literature, the future is tech-centric. Much of the basic computer programming will be done in low-cost countries: Any job in which the work product can be sent over the internet is at risk. The best shot at long-term remunerative tech employment in the U.S. will be for people with tech chops plus management, business, and/or domain expertise. For example, the SQL programmer able to successfully lead a remote team of water-supply experts, including understanding the organization’s business needs should continue to have excellent job prospects.
Also, machine learning/deep learning programmers, data security specialists, and especially people with expertise in both, should continue to find abundant job opportunities.
Direct health care providers. Because the U.S. is moving toward “covering” everyone, cost-control pressures are creating high demand for intermediate-level providers: Nurse practitioners and physician assistants rather than physicians, clinical social workers rather than psychotherapists, physical therapy assistants rather than physical therapists, nurse anesthetists rather than anesthesiologists.
Ultra-simple self-employment. Complex or capital-intensive businesses suffer from too many ways to fail, and competition is fierce in those spaces. Businesses more likely to succeed are simple, lower-status, and cost little to start and run. A few examples:
- Tutoring, whether for children or adults, for example, in technology.
- Health or fitness coaching.
- Blue-collar design and implementation: carpentry, landscape, plumbing, mobile car-repair. (Commercial and residential.)
- Trendy items sold near mass-transit terminals. An example is selling hot local sports teams' tee-shirts and hats from a cart or even walking with a tall stack of caps on the vendor’s head near a stadium parking lot on game day.
Even in a challenging world, those career areas should remain robust.
Nevertheless, it seems clear that there won't be enough decent-paying stable jobs to go around, so we'll need to encourage the masses to employ people, for example, as personal assistants, masseuse, manicurist, nanny, maid, and chef. I call that The Assistance Army. We may also need taxpayer-subsidized lifetime education to enable people to keep up with the accelerating pace of change.
Those may be inadequate, so we may also need a guaranteed basic income paid by heavily by successful corporations and wealthy individuals, although even if they were taxed at confiscatory rates, that's unlikely to be enough to provide the literally billions of people with an even tolerable standard of living, including housing, food, transportation, education, and health care. Society's main hope may be that technology's cost savings will sufficiently lower the cost of living. Certainly, most people will have to learn to live a less materialistic lifestyle and derive pleasure primarily from being productive, creative outlets, and relationships.
In the personal sphere, for us to compete with robots even in the near term, we may all have to up our game—hotheads, whiners, etc. beware. So it might not be a bad time for us to start working on ourselves—while we still have the chance.
An optimistic vision
An optimistic scenario is that job loss will not be as great as the above suggests. Indeed a study by the Organisation for Economic Cooperation and Development projects only a 9% job loss in the next two decades. An argument for modest job loss is that it's far easier to predict jobs that will be lost than jobs that haven't yet been invented yet. For example, no one predicted that Facebook in 2017 would be hiring thousands of human monitors to extirpate "objectionable content?"
A further reason for optimism is forwarded by MIT's Ted Autor. He believes that many job losses will create new ones. For example, ATMs make it less expensive to open new branches, each of which require people.
Continuing the optimistic scenario, even if job loss is significant, that could be more than compensated for by automation's so lowering the cost of living that even if we earn little income, we'll do okay and have lots of leisure time to boot.
Longer term, it's even possible that we'll be able to accomplish more of what we want by using gene therapy or a chip embedded in our brain—Research to make that happen is already being funded by the federal government.
Let's project a few decades from now. Lest you think that's too far in the future to worry about, remember that environmentalists are urging us to act now to avert a possible disaster that's 50 to 75 years away.
A pessimistic scenario is that there won't be enough well-paying jobs nor money from successful corporations to provide everyone with even basic housing, food, transportation and health care. That would cause greatly increasing crime and civil unrest.
And here is a scenario predicted by minority of credible scientists, notably Ray Kurzweil. They believe that in a few decades, we'll reach The Singularity, the point at which computers will be as intelligent as people. From then on, thanks to deep learning (the computers improving themselves based on past performance), computers' advantage over humans will grow and grow.
Some, like the aforementioned Stephen Hawking, Bill Gates, and Elon Musk worry that such a computer could calculate that it should take control over part or even the entire world. Is it beyond the pale that hyperintelligent, superpowerful computers will agree with environnmentalists that humans are destroying the planet and thus the computers cut electric power to the world, or even infuse our water supply with permanent birth control or even deadly poison, and thus kill off much of the world except the regions that are over-represented by environmentally responsible, highly intelligent, highly ethical people? Or could the computers decide the world would be better without people altogether?
And if that turns out to be the outcome of world evolution, then why would it only occur on Earth? After all, the universe is infinite, which would make it highly unlikely that Earth is the only planet that evolved that way. That in turn would mean that the universe contains many planets run by computers. It's even possible that the computers on some other planet were responsible for creating Earth and thus capable of destroying it.
The range of scenarios would seem circumscribed by these. How likely do you think each of these are?
The truth may well be something we can't even envision. After all, he who lives by the crystal ball usually eats broken glass.
I read a less comprehensive version of this essay on YouTube.