How Will Artificial Intelligence Affect Human Careers? Here Are 11 Forecasts

It's not being replaced that should worry you — it's having to keep learning new skills.

2017 is widely seen as the year artificial intelligence began drawing enormous attention worldwide. If you were paying attention, the news arrived almost daily: Alphabet, Google's parent company, testing self-driving cars in public; Alibaba announcing a 100-billion-yuan investment to found its DAMO Academy; Baidu's robots moving into KFC.

New technology always creates demand for new jobs and destroys others. The consulting firm Gartner has said that before 2019, AI will eliminate more jobs than it creates. But from 2020 on, the jobs AI creates will outnumber the ones it kills off — AI will "kill" work and make new work at the same time.

One extremely optimistic line of argument holds that people will get to do the work they love, efficiently, with more time for holidays and pleasure; that the entertainment industry will boom as never before; and that by then we will need new measures of human productivity.

That kind of talk feels like listening to Aldous Huxley narrate Brave New World. Still, the future is coming whether or not we like it, so we may as well look at what people are actually predicting.

Drawing on the new ten-year forecast recently published by the U.S. Bureau of Labor Statistics, Zhaopin's 2017 Report on Supply, Demand and Development in the AI Job Market, Michael Page (China)'s China Robotics and Artificial Intelligence Report, plus other reports, academic papers and press coverage, we have pulled together 11 points you may find interesting about how AI will change human work — mostly in the Chinese and American markets.

Pay in AI fields sits in the top 1% of the pyramid

According to the Global Artificial Intelligence Talent White Paper, published jointly by the Tencent Research Institute and BOSS Zhipin, there are about 300,000 AI professionals worldwide, while demand runs into the millions. Roughly 100,000 of them work in universities and 200,000 in industry. There are 367 universities worldwide with AI research programs, turning out about 20,000 AI graduates a year — nowhere near enough.

Chen Huijie, a regional director at Michael Page (China), told QDaily that pay competition for AI candidates sits at the top 1% of the pyramid. Many tech companies offer equity to top people, and some Chinese tech giants, hoping to attract talent from abroad, will let a candidate start work from anywhere in the world.

Two weeks ago Google opened a center in China devoted to artificial intelligence and began hiring for several AI and machine learning roles, including a technical lead for its machine learning research institute and a product manager for cloud machine learning. Microsoft and IBM are hiring Chinese staff in the same fields.

The available data shows that in China, new graduates in AI jobs start at roughly 12,500 to 25,000 yuan a month; three years out of school, an AI engineer can generally expect that pay to double. Competition is fierce enough that companies fighting over a candidate will offer raises of 30% to 50%.

Data analysts cross over most easily; people with machine learning skills are scarcest

A Zhaopin survey found that data analysis is the most mature skill among all AI talent, and the most common, at 48%. As the "infrastructure" of the field, data analysis gives job seekers the lowest barrier to moving into AI, and they are the group most likely to make the jump. Next come job seekers with image-processing skills, at 14%, and data mining, at 13%. Only 6% have machine learning skills.

AI talent clusters at a few companies; in China, Huawei has hired the most

Because AI talent is scarce, it is heavily concentrated at large companies — in practice, at a handful of tech giants. In China, most of the existing talent sits at Huawei, Lenovo and BAT, with Huawei holding the largest share, 30%. Worldwide, more than 50% of the industry's top AI experts have already been signed up by Google.

Wang Xiaoyu, chief scientist and VP of AI at Intellifusion, a Shenzhen company working on visual AI, said: "In the short term, hiring AI talent is very expensive for a startup. When we hire a computer vision PhD fresh out of school in the U.S., the salary is $300,000 to $400,000 a year."

AI talent is concentrated in first-tier cities. In Beijing, Shanghai and Shenzhen, the supply of talent exceeds the share of demand; in Guangzhou and Hangzhou, supply falls short of demand, so the shortage there is worse.

No Chinese university stands out as especially good at training AI talent. The three schools with the largest shares of graduates now in the job market are Tsinghua University, Xidian University and Huazhong University of Science and Technology.

As "unmanned stores" catch on, retail work will shrink sharply

Almost every report on artificial intelligence makes the same point: future stores will rely more on machines and less on human staff, and shoppers increasingly prefer to pick, pay and even bag their own goods. Over the past year, experiments with "unmanned stores" have multiplied — Walmart's unmanned supermarket project aimed at Amazon, Japanese convenience stores responding to an aging population, cashless stores in China.

The World Economic Forum has estimated that once all the automation technology now available is put to use, 30% to 50% of retail jobs worldwide will be at risk of disappearing.

The clearest example is the self-service machine, in wide use for more than 20 years. The British consulting firm Retail Banking Research calculates that the number of self-checkout machines worldwide will rise from 200,000 in 2013 to 325,000 in 2021. Over the next decade, U.S. grocery stores alone will shed 42,700 cashiers and 28,800 baggers.

Self-driving technology will dent demand for drivers far less than Uber and Didi have

Last October, Uber began letting self-driving trucks take orders; in China, driverless buses are already in service. Safer, more efficient autonomous driving suits long-haul freight, especially as e-commerce keeps growing fast.

Some forecasts say 80% of American trucking jobs will vanish — but like many such reports, this one gives no date.

Economists at the U.S. Bureau of Labor Statistics do not think self-driving trucks will replace drivers that quickly. They forecast that heavy-truck driving jobs will actually grow by 114,000 over the next decade.

Teri Morisi, who directs the agency's occupational outlook division, said: "We think autonomous driving is still a bit further off than our forecasts for the trucking industry account for. Advances like platooning and brake assistance will make truck driving safer and more energy-efficient, but they won't change [market] demand for truck driving."

As for other driving jobs, the number of bus drivers in American urban transit systems will rise 8.9%, while the number of taxi drivers will fall 41.2% — though that has more to do with Uber and Lyft.

85% of North America's lost manufacturing jobs went to "machines"

The prophecy of Charlie Chaplin's Modern Times is coming true, minus the irony.

According to the latest ten-year forecast from the U.S. Bureau of Labor Statistics, 14 of the 20 industries losing jobs fastest are in manufacturing. Apart from tobacco, which is being hit hard by falling demand, every one of them is tied to the spread of robots.

Between 2000 and 2010, the United States and Canada lost 5.6 million manufacturing jobs. Of those, 85% were lost to machines; only 13% went to developing countries with cheaper labor.

Blue-collar workers are being replaced because their work is already mechanical.

The Boston Consulting Group (BCG) argued in 2015 that industrial robots are becoming a better and better deal: the average cost of buying and running one will drop from $182,000 in 2005 to $103,000. By 2025, annual spending on industrial robots in the world's 25 largest exporting countries, including the United States, China and South Korea, will be growing 10% a year.

Accounting and auditing, long written off, will keep growing for another decade

In a widely cited 2013 paper, The Future of Employment: How Susceptible Are Jobs to Computerisation?, the Oxford professors Benedikt Frey and Michael A. Osborne found that "accounting and auditing" was among the most endangered occupations, with a 94% chance of being automated: processing payments, auditing the books and similar work could well be handed to software.

The good news is that it should not happen in the next ten years. Bureau of Labor Statistics forecasts show jobs in the American accounting industry rising from 985,700 this year to 1,063,600 in 2026. "Accounting and auditing," about 36% of the industry's jobs, will grow 11.8% — one of the fastest-growing occupations in the country.

The least skilled jobs in accounting are another matter: typists, data entry clerks, switchboard operators and file clerks. In a paperless office, every one of these will shrink by more than 10%, with typists down 32.9%.

That decline is not confined to accounting; it runs through every industry. At the end of their paper, the two professors listed the odds that machines would take over 702 different occupations. Telemarketing and data entry are the likeliest of all, at 99%.

ATMs keep getting smarter: one machine, one branch

"ATMs can already do 90% of what people can do," said Andy Mattes, CEO of Diebold, the world's largest ATM maker. As cash falls out of use, ATM companies are working hard to make their machines perform better and think more.

The Future of Employment put the odds of a bank teller's job being automated at 98%. Over the next decade, the number of tellers at U.S. depository institutions will fall from 444,000 to 390,000, down 12.1%.

Mechanical work — collecting customer information, checking loan documents, drawing up contracts — will be automated. A McKinsey report found that these tasks, which eat up 43% of the time of employees at financial institutions, can all be done by machines. That is the practical advantage of even weak AI.

Japan's largest bank, Bank of Tokyo-Mitsubishi UFJ, has announced that it will hand clerical work equivalent to 9,500 employees' workload over to software, but says it will not cut staff as a result; instead, employees will be asked to spend more of their time on the bank's wealthiest clients.

When buying a home, take a human's advice

Somewhat surprisingly, real estate agents — a service job — have also been listed at times among the occupations most at risk from automation. U.S. Bureau of Labor Statistics data shows jobs in real estate falling 12.1% over the next decade, with both sales representative and agent positions down 7.9%.

As with salespeople, the unpleasantness of buying a home has pushed buyers toward posting and browsing listings online themselves. The real crisis for agents, it seems, is the quality of their service.

Artificial intelligence can help a buyer weigh loan options and narrow the field, which lightens an agent's workload.

But for now these jobs are unlikely to be replaced outright. At the final, mortgage stage of a purchase, buyers still lean on human advice: the most trusted sources are agents (41%) and banks (35%), and only 11% of people are willing to make the decision on a machine's opinion.

The AI era will need more jobs that supply emotion and creativity

Once artificial intelligence can handle more routine work, human energy and intelligence will be free for the things people are good at and machines are not: making things of the spirit, art and culture. Recreational therapists, for instance, have only a 0.28% chance of being replaced.

Zhaopin also screened the jobs that call for creativity and emotion, and found that where thought, spirit and invention lead the way, demand is shifting toward what only a person can give. A therapist, for example, has to treat a patient's particular needs through feeling and psychological insight, not a simple personality test; copywriters and creative staff in advertising are being asked to let loose more human imagination, rather than to operate design software.

But even in these fields, the jobs that require only basic skills will go. Luban, the "AI designer" that appeared at this year's Double 11 shopping festival, is a ready example: by learning and practicing without pause, it mastered every formula and style of banner design, and can turn out 8,000 posters a second.

So human designers will contribute more thinking, and less practiced repetition.

Kai-Fu Lee, now president of the Sinovation Ventures AI Institute, predicts that the talent pyramid of the future will hold roughly four groups: first, the leaders of innovation; second, the top professionals in every field, whether medicine, journalism or anything else, who will be a small minority; third, people in the humanities and the arts, whom AI cannot replace; and fourth, workers in services.

And according to a report in The New York Times, artificial intelligence is already at work in music, painting and other fields — creating not only new kinds of art, but new artists.

It's not being replaced that should worry you — it's having to keep learning new skills

The fight for AI talent is a global one. The United States has the largest supply; Japan, India, Israel, Britain, France, Germany and China all have heavy demand as well.

For the many companies that cannot find AI people, hiring from other industries, or hiring engineers who have switched fields, has become the second-best option, and many large companies have set up in-house training programs.

Graduates in physics, digital technology, electronics and other subjects have all "crossed over" into artificial intelligence. The technical barrier to changing which industry a technology is applied in is often not that high: with basic math and some knowledge of the field, crossing over is possible.

As noted above, people working in big data will find it easiest to cross over. But the higher-end demand — machine learning, image processing and other jobs with a steeper technical barrier — is hard to fill that way, and companies rely more on recruiting abroad.

The crisis of people being replaced by machines will not really arrive in the next few years. Being chased by machines into learning new skills and keeping up with what the workplace wants, though, is already happening.

Cover image: Blade Runner 2049