“You can’t create a monster, then whine when it stomps a few buildings.”
Yeardley Smith
I imagine the high-and-mighty dinosaurs never saw it coming: the extinction event—possibly triggered by an asteroid—that wiped them out roughly sixty-six million years ago.
Similarly, fifty-six-year-old Georgia resident Sherry Johnson never saw it coming: the robots that put her out of work, not once, but three times. First, a robot claimed her job at a local newspaper, where she laid out pages and fed paper stock into printing machines. Second, a robot replaced her at a factory that produces breathing machines. Third, machines took over her job doing inventory and filing. “It actually kind of ticked me off,” she says, “because it’s like, how are we supposed to make a living?”377
There’s a lively debate today about what impact the fourth industrial revolution—artificial intelligence and robots—will have on the human job market. Will it devastate it or invigorate it?
Predictably, the hypesters see nothing but blue skies. “Historically, technology has created more jobs than it destroys and there is no reason to think otherwise in this case,” says Vinton Cerf, Google’s Chief Internet Evangelist. (Yes, that’s his actual title.) “Someone has to make and service all these advanced devices.”378
“The historical record provides strong support for this view,” agrees Jerry Kaplan, Stanford professor and author of Artificial Intelligence: What Everyone Needs to Know. “After all, despite centuries of progress in automation and recurrent warnings of a jobless future, total employment has continued to increase relentlessly . . .”379
McKinsey Global Institute (MGI) offers a striking historical example of such job creation in its December 2017, report, “Jobs Lost, Jobs Gained: Workforce Transitions in a Time of Automation.” The invention of the computer, it points out, killed some 3.5 million jobs in the United States, such as typists, secretaries, typewriter makers and repairpersons, bookkeepers, and auditing clerks. But it minted 19.3 million new positions, in computer manufacturing and supplies, IT system administrators, software engineers—all of which gave rise to the tech giants of Silicon Valley and beyond: Oracle, IBM, Microsoft, Google, Apple, and Facebook, to name just a few.380
Today, even as the AI revolution rampages through the traditional workplace, there is a soaring demand for workers with technical educations, such as might be gotten at a community college or trade school.381
“[L]asting job creation will require an understanding of important new dynamics in the global labor market,” says IBM CEO Ginni Rometty. “This is not about white collar vs. blue collar jobs, but about the ‘new collar’ jobs that employers in many industries demand, but which remain largely unfilled.”382 Still, the AI revolution is completely without precedent, cautions Jeremy Howard, AI pioneer, entrepreneur, and Distinguished Scholar at the University of San Francisco. “People aren’t scared enough, you know. Far too many people are sounding like climate change denialists. They’re saying, ‘Don’t worry about it, there will always be more jobs.’ And it’s founded on this purely historical thing of like, ‘Oh, there’s been a revolution before . . . and after it there was still enough jobs,” he says. “But it’s a ludicrously short-sighted and meaningless argument, which incredibly smart people are making.”383
Whereas in past revolutions steam power replaced horsepower, electricity replaced mechanical gears and gaslight, and computers replaced typewriters, the AI revolution is specifically designed to replace us.
Yes, the invention of computers did create millions of opportunities to service and supervise the new machines. But AI is being created to service and supervise itself. That’s the whole point of AI—to replace us with something superior.
Eventually, AI will replicate better versions of even our prized intuition and other human je ne sais quoi. Indeed, as we’ll see in this chapter, it’s well on its way to doing just that.
“We are approaching a time when machines will be able to outperform humans at almost any task,” says Moshe Vardi, director of Rice University’s Ken Kennedy Institute for Information Technology. “I believe that society needs to confront this question before it is upon us: if machines are capable of doing almost any work humans can do, what will humans do?”384
Actual Job Losses
AI’s assault on human employment arguably began back in 1961, when General Motors in Trenton, New Jersey, hired the world’s first digital, programmable robotic arm—Unimate—to handle hot-metal die castings and perform precision spot-welding on car bodies. Now on display at the National Museum of American History in Washington, DC, Unimate inspired the design of robotic arms that presently dominate assembly lines all over the world.385
Admittedly, robots and AI comprise just one reason US manufacturing jobs have taken a huge hit in recent years. Other culprits include shifting trade imbalances, outsourcing, and fluid product demands, so it’s difficult to get reliable statistics on the actual damage AI is inflicting on the human worker.
Nevertheless, Ball State University researchers offer us some ominous clues in their 2017 study of what happened from 2000 to 2010, “which was the largest decline in manufacturing employment in US history.”386 During that fateful decade, plant productivity soared—not because human employees suddenly became more efficient, the report explains, but because of “automation and information technology advances.” Altogether, robots and artificial intelligence accounted for 88 percent of the historic spike in productivity.
“Had we kept 2000-levels of productivity and applied them to 2010-levels of production,” the researchers report, “we would have required 20.9 million manufacturing workers. Instead, we employed only 12.1 million.” The difference was taken up by smart robots. Economists at MIT and Boston University came to a similar conclusion after analyzing the effect of robot employment in the United States between 1990 and 2007. In their study for the National Bureau of Economic Research, they controlled for factors such as “imports from China and Mexico, the decline of routine jobs, offshoring, other types of IT capital, and the total capital stock.” According to the study’s authors, Daron Acemoglu and Pascual Restrepo, “one more robot per thousand workers reduces the employment-to-population ratio by about 0.18–0.34 percentage points and wages by 0.25–0.5 percent.” Translated it means that, on average, employing one industrial robot results in un-employing 5.6 humans—more so men than women—and reducing wages.387
And don’t for a minute think the trend spells trouble solely for repetitive, blue-collar jobs; pink- and white-collar workers are equally in danger. “We are just seeing the tip of the iceberg,” warns Sebastian Thrun, the celebrated guru of AI. “No office job is safe.”388
“We often think about automation as applying to front-line, low-wage, low-skill activities and jobs,” avers Michael Chui, a computer scientist with MGI. But that’s only the beginning. “Almost every job in the economy has a significant percentage of activities that can be automated.”389
Indeed, AI software currently replacing white-collar jobs is much cheaper to buy than a typical industrial robot. So it could be said the fourth revolution poses a greater threat to paralegals and accountants, say, than the average manual laborer.
Here are some specific examples of human tasks that are in danger of going the way of the dinosaurs:
Kinema Pick—made by Kinema Systems—is a robot arm that unloads tightly packed, oddly shaped, oddly colored boxes from a pallet and places each one gingerly onto a conveyor belt. It sounds simple, but in fact it is a complicated task requiring a strong arm, sharp eye, and keen mind that only humans have been able to do—until now.
“With our system, it’s self-training,” explains Sachin Chitta, CEO of Kinema Systems. “It starts from scratch, having no idea what a box is, and as it picks, it’s training itself. Once it picks a box for the first time, it builds a model of what the box looks like and uses that to speed up its pick the next time.”390
Once it gets going, says Chitta, Kinema Pick outpaces even the most experienced human longshoremen on the planet. Also, it doesn’t require a break every several hours and will never file an injury claim.
Construction
SAM—short for “semi-automated mason,” created by Construction Robotics—can lay up to 3,000 bricks a day, which beats the pants off the average human mason’s 500 bricks a day. It runs on tracks, slathers on the mortar, and places each brick or half-brick exactly where it needs to be, using a laser system for precision guidance. Humans are still needed, but are reduced to tending to Sam, like so many underlings, keeping the robot fed with bricks and mortar, smoothing the joints along the way, and laying any bricks requiring odd angles. “Once it starts running,” says Scott Peters, the company’s CEO, “it can run continuously within a given setup for hours on end.”391
Retail Manufacturing
Right now, sports shoes are mostly handmade; but that’s changing. In 2015 Adidas began making sneakers in a highly automated, futuristic-looking facility in Ansbach, Germany, called the Speedfactory. In late 2017, they opened a second Speedfactory in Acworth, Georgia, just outside of Atlanta.392
Adidas is automating to capitalize on the explosive, fast-paced made-to-order/buy-now/wear-now phenomenon created by online shopping. A research report by Morgan Stanley, which sees automation as the future of retailing, explains the phenomenon this way:
“Let’s say the consumer likes one style, but just wants to change one thing. He’ll be able to do that. What if he wants the product customized to his foot? No problem. Let’s say he’s traveling and wants to pick it up at a Foot Locker store near his hotel the next day? Easy.”393
Farming
Facing an increasing shortage of field hands, many farmers are embracing AI in a big way. “The availability of labor is a growing concern,” says Kenneth Parker, executive director of the Florida Strawberry Growers Association. “The strawberry growers’ margins have become incredibly thin. When you have such thin margins, any additional expense is unbearable.”
Enter a tireless strawberry-picking robot, created by Harvest CROO Robotics and due out in 2018.394 Its keen “eyes” can spot ripe berries; its “hand” pushes aside leaves; and its wheel of plastic “fingers” gently pluck the berries. The picked berries go into a holding tank, where the robot inspects them one last time before packing them into individual retail containers.
Harvest CROO Robotics predicts a single robot will replace more than thirty human pickers.395 Moreover, it claims a single person will be able to keep track of all the robots working in a field and to make any needed repairs.
“I haven’t talked to one strawberry grower in Florida or California that said he wouldn’t go to a machine,” says Gary Wishnatzki, the company’s co-founder. “Any major crop that requires hand harvesting is going to be mechanized in the next 20 years.”396
In a dramatic case of life imitating art, the Dubai police force has hired a real-life Robocop—the first of its kind in the Middle East—to help patrol its busy, urban areas. Dressed in a faux police uniform (snappy cap and all), the life-size robot moves around on wheels, shakes hands, salutes, can recognize faces, read license plates, and alert his fellow officers to suspicious-looking packages and situations. It also sports a touch screen on its chest, which people can use to report a crime or look up a speeding ticket.
The aim is for robots to make up one-fourth of the Dubai police force by 2030. “These kind of robots can work 24/7,” remarks Brigadier Khalid Nasser Al Razooqi. “They won’t ask you for leave, sick leave or maternity leave. It can work around the clock.”397
For the record, a robot cop named K9—created by Knightscope398 and hired by San Francisco’s SPCA to guard its animal shelter, which is in a high-crime neighborhood—did not fare so well. According to one newspaper report, the five-foot-tall K9 “was battered with barbecue sauce, allegedly smeared with feces, covered by a tarp and nearly toppled by an attacker.” Worst of all, K9 was accused of terrorizing homeless people in the area.399
In December 2017, San Francisco’s SPCA was forced to fire K9. “We piloted the robot program in an effort to improve the security around our campus and to create a safe atmosphere for staff, volunteers, clients and animals,” laments Jennifer Scarlett, president of the venerable nonprofit. “Clearly, it backfired.”400
Ground Transportation
A score of companies—from Alphabet to Ford to General Motors— have invested about $80 billion in driverless cars and are road-testing them in cities such as San Francisco, Boston, Detroit, and Scottsdale. They’re competing for the lion’s share of what they see as the greatest money-making opportunity since cars replaced horses and buggies.
In the next few years alone, it’s widely expected tens of thousands of the so-called headless cars will be plying our roads. “I think in twenty years in major cities, fifty percent of cars will be driverless,” predicts John Quelch, marketing professor at Harvard Business School.
Waymo—Alphabet’s driverless car subsidiary—is aggressively developing a self-driving Chrysler Pacifica minivan that will compete with Uber, Lyft, and traditional taxis. In 2017, using AI and a suite of lasers, radar, and cameras to “see” and “think,” the minivan successfully maneuvered around a bicyclist, pedestrians, a roundabout, construction-zones, and merging traffic.
“Our intention, make no mistake, is to go fully driverless and let the public access this technology on public roads,” says Waymo’s CEO, John Krafcik. “We’re getting to the point now where, I think it’s fair to say, we’re really close.”401
Part of the drive behind driverless cars comes from the tragic fact that US road deaths are rising. In 2016 more than 40,000 people were killed in crashes, the vast majority of which were caused by human error.402
But what about inevitable computer errors? In 2016 a self-driving Tesla S saloon car rammed into a tractor-trailer because its computer vision and artificial intelligence were unable to make out the truck’s white exterior against a sunny sky.
Worse yet, in Tempe, Arizona, on March 18, 2018, a driverless Volvo XC90 sport utility vehicle operated by Uber hit and killed forty-nine-year-old Elaine Herzberg. She was walking her bike across the street at night and the driverless car, going forty miles per hour in a forty-five– mile-per-hour speed zone, failed to see her.403
And what about the split-second, life-and-death decisions driverless cars will need to make, even when they do spot trouble ahead? For example, what does a headless car do if one morning an oncoming truck suddenly crosses into its lane while driving past a mother and small child waiting for the school bus?
Sebastian Thrun, a highly respected engineer who launched Google’s self-driving project, thinks he knows the answer. In a video interview, he tells a Bloomberg reporter—with body language that says it’s the simplest problem in the world to solve—that “If it happens where there is a situation where a car couldn’t escape, it’ll go for the smaller thing.”404
Yikes—as in the small child?
Thrun’s smug but trite thinking on the subject shows how very, very far we have to go before driverless cars can be trusted over human drivers, even with our imperfections. “As human beings, we have hundreds of thousands of years of moral, ethical, religious and social behaviors programmed inside of us,” observes Frank Menchaca, chief product officer at SAE International, formerly Society of Automotive Engineers. “It’s very hard to replicate that.”405
Food Service Industry
In 2017—as part of an initiative called “Experience of the Future”— McDonald’s installed self-service kiosks in 2,500 of its locations to replace the front-counter folks who greet us and take our orders. Wendy’s did likewise in 1,000 of its restaurants.
At the start of 2018, California-based Jack in the Box said it might do the same thing—replace cashiers with self-serve kiosks—as a way to cope with the rising minimum wage in the Golden State and elsewhere. “As we see the rising costs of labor,” says CEO Leonard Comma, “it just makes sense.”406
CaliBurger—the worldwide chain touting fresh, California-style burgers—announced it will replace its kitchen staff with robots named Flippy in at least fifty locations.407 It began doing so in March 2018 by installing Flippy in its Pasadena, California, store.408
Flippy—whose creators at Miso Robotics call a “kitchen assistant”— is versatile and deeply intelligent. According to David Zito, Miso’s CEO, Flippy does tasks that involve “the high pain points in restaurants and food prep. That’s the dull, dirty and dangerous work around the grill, the fryer, and other prep work like chopping onions.”409
Those are the very jobs, however, that young people rely on to get their foot into the labor market. Last summer, my own teenage son worked at a popular fast-food restaurant doing the very things Flippy will be commandeering.
Pushing the frontiers of automation even further, Momentum Machines in San Francisco has created a burger-bot it claims can do everything—i.e., make a gourmet burger from scratch and even package it.410 “The burgers sold at 680 Folsom [the location of Momentum’s proposed restaurant] will be fresh-ground and grilled to order, served on toasted brioche, and accented by an infinitely personalizable variety of fresh produce, seasonings, and sauces,” boasts a Craigslist ad posted by the company.411
“Our device isn’t meant to make employees more efficient,” declares Alexandros Vardakostas, the company’s brash CEO. “It’s meant to completely obviate them.”412
Retail Stores
In a trial program at eleven Lowe’s stores throughout San Francisco, LoweBots are greeting customers, helping them find products, and taking inventory. The five-foot-tall, roving robots actually walk customers to the items being sought and can be programmed to understand twenty-five languages. Moreover, their AI is designed to detect complex patterns in the buying behavior of a store’s customers, helping Lowe’s keep the right kind and number of products in stock at all times.413
Similarly, Walmart is employing robotic store clerks in fifty of its stores in Arkansas. The brainy, two-feet-high roving machines will use artificial vision to scan store shelves, taking inventory, checking pricing, and looking out for misplaced items. Walmart claims by delegating these “repeatable, predictable, and manual” tasks to a machine, its human employees will be able to devote more quality time to customers.414
Not to be left out, in January 2018, Amazon opened an 1,800-square-foot, artificially intelligent, completely cashier-less grocery store in downtown Seattle, called Amazon Go. Using computer vision, storewide sensors, and a massive, deep-learning computer program, the store enables customers merely to swipe an Amazon Go app as they enter, browse the shelves—for now, stocked with freshly made breakfast, lunch, and dinner items, as well as some essentials—grab whatever they want and then simply walk out. No cashiers to deal with. As customers exit, the store automatically senses and tallies the items they’ve selected and directly charges their Amazon account.415
News Industry
Even news people, such as myself, are now in the crosshairs. The AP, Yahoo! Sports, Comcast, and other major media companies have hired an artificial journalist program called WordSmith—created by Automated Insights—to write millions of articles a week on topics that are heavy on statistics, such as fantasy football and quarterly earnings. WordSmith pumps out as many as 2,000 articles per second and credits itself with the byline: “This story was generated by Automated Insights.”
Both the AP and Automated Insights swear no jobs have been lost to WordSmith—not yet, anyway—but the AP is reportedly looking to expand its employment of robotic journalists.416
On the TV side of the news business, Hiroshi Ishiguro, director of Osaka University’s Intelligent Robotics Laboratory, has high hopes for his humanlike robot, Erica. “We’re going to replace one of the newscasters [in Japan] with the android,” he says, perhaps as soon as April 2018. Some say Erica will scarcely be more than a glorified talking puppet; but then, honestly, the same could be said of many high-priced human TV anchors.417
Financial Services
Accenture, the worldwide consulting company, predicts a whopping 80 percent of all traditional finance services will soon be automated, cloud based, and handled by decidedly nontraditional personnel.418
Case in point: Pilot Travel Centers, which operates the Pilot Flying J truck stops, used to employ eighty fulltime bookkeepers. “It was just awful. There were humans everywhere,” recalls David Clothier, the company’s treasurer. Today, thanks to an AI program, Pilot employs only ten fulltime human bookkeepers.419
Similarly, the insurance industry is using AI to replace claims agents. In 2017 Japan’s Fukoku Mutual Life Insurance purchased an AI program from IBM, which “can analyze and interpret all of your data, including unstructured text, images, audio and video.” As a result, Fukoku was able to downsize its claims department from 131 human adjusters to just ninety-seven.420
Marketing
In 2016 Cosabella, the famous woman’s lingerie retailer, fired its digital advertising agency and hired Albert, an AI program. After the first month, Albert increased Cosabella’s return-on-ad-spend (ROAS) by 50 percent and after three months, 336%.
“Albert’s success driving conversions on Facebook resulted in a 2,000% increase in purchases originating from the channel,” raves Courtney Connell, Cosabella’s marketing director, adding: “He doesn’t sleep, he’s fast, he doesn’t get into a fight with his girlfriend and lose focus.” For Connell, there’s no going back. “After seeing Albert handle our paid search and social media marketing, I would never have a human do this again.”421
Legal Services
JPMorgan Chase has recruited an AI program called COIN (Contract Intelligence) to review and interpret 12,000 new commercial loan agreements every year—work that used to be done by human lawyers. The result is fewer costly blunders in servicing the loans, most of which were caused by human error. COIN quickly does the work of more than 170 fulltime attorneys—a total of 360,000 hours of legal work annually.422
Stock Trading
Automated stock brokers and investment managers capable of making rapid-fire, split-second decisions are increasingly rattling Wall Street. On February 5, 2018, these computerized bots were blamed for driving down the Dow Jones average by 800 points in just ten minutes and a whopping 1,597 points within a single day—the worst such drop in trading history.
“The computers react to evidence exponentially faster than any human—think millionths of a second, instead of minutes—and can move en masse, trading at high volumes around the world,” reported The Washington Post. “That makes them potentially more dangerous.”423
David Weild IV, former vice chairman of Nasdaq, agrees, lamenting: “We’ve created a stock market that moves too darn fast for human beings.”424
Now, there’s every indication things are about to get even faster and more dangerous.
In late 2017 EquBot—a San Francisco-based startup company— teamed up with Watson, IBM’s legendary cognitive supercomputer, to produce the first AI-powered stock bot. Its superhuman capabilities include reading and analyzing more than one million pieces of market-relevant information a day, such as daily news stories and earnings reports, as well as economic, consumer, and industry trends. It then uses the up-to-date information to make predictions about the stock prices of 6,000 publicly traded companies and alerts investors large and small about the optimal times to buy and sell.425
“EquBot AI Technology with Watson has the ability to mimic an army of equity research analysts working around the clock, 365 days a year, while removing human error and bias from the process,” boasts Chida Khatua, EquBot’s CEO.426
And, oh yes. Watson only gets smarter with age.
Lip Reading
LipNet, a lip-reading program out of the University of Oxford, scores a 93 percent accuracy rate, compared with only 52 percent for expert human lipreaders. “LipNet aims to help those who are hard of hearing,” says Yannis Assael, one of the researchers, adding “it has the potential to revolutionize speech recognition.”427
Medical Services
“Artificial intelligence and robotics spell massive changes to the world of work,” says Matt Beane, project scientist at the University of California, Santa Barbara, including the medical profession. “Over a third of U.S. hospitals have at least one surgical robot.”428
The “800-pound gorilla” of robotic surgery, Beane says, is the da Vinci robot surgeon, made by Intuitive Surgical, in Sunnyvale, California. “This is a four-armed robot that holds sticklike surgical instruments, controlled by a surgeon sitting at a console 15 or so feet away from the patient.”429
“The da Vinci System features a magnified 3D high-definition vision system and tiny wristed instruments that bend and rotate far greater than the human hand,” explains the company website. “As a result, da Vinci enables your surgeon to operate with enhanced vision, precision and control.”430
HeroSurg—a laparoscopic robot surgeon being developed at Harvard and Australia’s Deakin University—lets human operators actually feel what’s going on. So, for example, “if they’re cutting the tissue,” says project leader Mohsen Moradi Dalvand, “they can feel the amount of force they are applying to the tissue.”431
“The major drawback of the current [robotic systems] . . . is the lack of tactile feedback,” remarks Suren Krishnan, a surgeon at Royal Adelaide Hospital who collaborated in HeroSurg’s creation. “Tactile feedback allows a surgeon to differentiate between tissues and to ‘feel’ delicate tissues weakened by infection or inflammation and dissect them more carefully.”432
Yet another AI surgeon is Cyberknife, a robotic radiation oncologist developed at Stanford University that delivers radiation treatments within 0.5 millimeters of its target. That’s a precision far better than any human oncologist could possibly achieve. Moreover, Cyberknife’s x-ray vision and AI enable it to constantly adjust for a patient’s movements, so it can zap tumors with pinpoint accuracy, without the patient being clamped down.433
AI is also targeting other medical professionals, not just surgeons.
Watson, for one—IBM’s resident AI smarty pants, famous for whooping the all-time Jeopardy! champs [see ROBOT: MEMORY LANE]—is now after radiologists, a seriously beleaguered lot. There are upward of 40,000 of them in the United States, and they must scrutinize more than sixty billion images per year, which include ultrasounds, MRIs, x-rays, CT scans, PET scans, and so forth.434 That means each radiologist must read about 1.5 million images per year or about one image every twenty-one seconds.
After studying thirty-plus billion medical images at the feet of some of the finest radiologists in the world, Watson is now offering to help these overworked humans. Not only does Watson scan images for potential signs of disease, it takes into consideration (1) everything there is to know about patients, based on a careful reading of their electronic medical records; (2) information about other patients with similar profiles; and (3) the very latest medical research.435
IBM has high hopes Watson will “help doctors address breast, lung, and other cancers; diabetes; eye health; brain disease; and heart disease and related conditions, such as stroke.”436 But already there are signs Watson is likely to go from being a helpful collaborator to being a better, less expensive replacement—especially given the US radiologist’s median annual salary of more than $500,000.437
In a recent head-to-head experiment, for example, Watson and eight expert human doctors were tasked with searching 100 skin images for signs of melanoma. Watson’s accuracy rate beat the doctors’ by 76 percent to 70.5 percent.438
Entertainment
Hollywood, too, is seeing the writing on the wall, with a growing number of filmmakers using AI to save time and money. Peter Jackson, for example, used Massive, a sophisticated AI program, to generate the sprawling, lifelike armies in The Lord of the Rings movies.
It’s only the beginning.
“At some point,” predicts Stephen Regelous, Massive’s creator, “you’ll be able to create an actor that doesn’t know he’s not real.” And not just human actors are in jeopardy. Script supervisors, editors, and CG artists are all at risk of being upstaged by AI. “It’s all over by 2045,” warns Regelous. At that point, “we [humans] are no longer running the show.”439
Scientists at Rutgers University had art experts and laypeople evaluate paintings by accomplished human artists alongside paintings generated by AI—without telling them which was which. Amazingly, the judges could not tell the two apart. They “thought that the generated images were art made by an artist seventy-five percent of the time,” reports Ahmed Elgammal, director of Rutger’s Art and Artificial Intelligence Laboratory.440
The scientists also asked the judges to rate the paintings’ aesthetics— “to rate the degree they find the works of art to be intentional, having visual structure, communicative, and inspirational.” To the researchers’ surprise, the judges rated the AI paintings higher than the real deals.441
At Magenta—a division within Google’s AI lab—scientists have created NSyth, an AI musician that can generate instrumental sounds never heard before—for instance, one 17 percent piano and 83 percent trumpet.442
Engineers in other labs are also creating artificial musicians. For instance, ones that study the repertoires of famous musicians, searching for subtle, stylistic patterns. These AI musicians are expected one day to create a Queen-like pop song, a Bach-like concerto, or any piece of music in the style of whatever artist you wish.
“This work has exploded over the last few years,” remarks artist Adam Ferris. “This is a totally new aesthetic.”443
Ministry
AI is even taking on certain heavenly tasks. In celebration of the 500th anniversary of the Protestant Reformation, BlessU-2, a robo-pastor in Martin Luther’s hometown of Wittenberg, Germany, pronounces blessings on worshippers. It’s the experimental handiwork of the Protestant Church in Germany, led by clergyman Stephan Krebs. “We wanted people to consider if it is possible to be blessed by a machine.” BlessU-2 has a touch screen on its chest that enables petitioners to select one of eight languages, a male or female voice, and a specific blessing—which the robot pronounces with upraised arms and divinely illuminated hands.444
Not to be outdone, Pepper—a robot created by Japan’s Softbank— has been ordained into the Buddhist priesthood to preside over funerals. No one has hired him yet, but he’s programmed and ready to deliver rites at a cut-rate price. A human priest comes with a $2,200 price tag; Pepper will officiate a funeral service for a mere $450.445
AI Engineering
In the greatest of ironies, Jeff Dean, a senior engineer at Google, has created AutoML, an artificial engineer designed to eventually do Dean’s job: design AI systems—maybe even better than Dean.446
Efforts to build artificial engineers are mushrooming, actually, because there is a severe worldwide shortage of human AI engineers. According to Element AI, a Canadian consulting firm, fewer than 10,000 people worldwide currently have the mathematics-intensive skills necessary to lead complex, high-level AI research.447 By contrast, Tencent, China’s tech giant, claims the number is more like 300,000.448
Either way, the number of qualified AI experts falls way short of the millions needed to keep pace with the exponential growth of the AI industry. Total investment in AI grew from $589 million in 2012 to $9.5 billion in the first five months of 2017.449
“Just as electricity transformed almost everything 100 years ago,” marvels Andrew Ng, a computer scientist at Stanford and the former chief scientist at Baidu, China’s version of Google, “today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years.”450
For artificial engineers such as AutoML, it certainly spells job security for many, many years to come.
According to a study out of England’s prestigious Oxford Martin School, 47 percent of all American jobs are at high risk of being computerized in the next decade or two.451 Scientists at MGI have reached a similar conclusion. “Our scenarios suggest that half of today’s work activities could be automated by 2055,” they say, “but this could happen up to 20 years earlier or later depending on various factors . . .”452
AI’s takeover of the workplace will force many people to switch occupations to stay gainfully employed. Doing that will require retraining.
But very few companies are helping employees in AI’s crosshairs prepare for the loss of their current positions. “Right now, only three percent of the executives we’re surveying,” reports Julie Sweet, CEO of the North American division of Accenture, a global management consulting firm, “are going to significantly increase their investment in upskilling their people.”453
In an interview at the 2018 World Economic Forum in Davos, Switzerland, Bill Gates appears to be fairly blasé about all this, illustrating how far hypesters will go to paint a rosy picture of the AI revolution. “Certainly, we can all look forward to the idea that vacations will be longer at some point,” he said cheerily. “The purpose of humanity is not just to sit behind a counter and sell things. You know, more free time is not a terrible thing.”454
Others are very worried that today’s average workers—people who, unlike Gates, can’t afford lots of extra free time or swanky retreats in the Swiss Alps—risk a future far from cheery. “Within thirty years, half of humanity won’t have a job,” cautions Antonio García Martínez, former Facebook executive and author of Chaos Monkeys: Obscene Fortune and Random Failure in Silicon Valley. “It could get ugly. There could be a revolution.”455