AI will do spectacular things but also will spectacularly disrupt labor markets around the world, including in the United States. Some of the direst predictions in recent years have been tempered, and while no sector will be unaffected, the effects will vary a lot across regions and economic sectors. Beyond jobs, AI raises deep questions about how humans think, and even about what it means to be human.
By Gregory F. Treverton
NOTE: The views expressed here are those of the author and do not necessarily represent or reflect the views of SMA, Inc.
Exposure to AI
One recent study concluded that a quarter of American jobs—36 million in 2016—will be highly exposed to automation in the coming decades, where “highly exposed” means more than 70% of tasks at risk of being substituted by machines. Thirty-six percent of jobs will be face medium exposure, and 39 percent low exposure. Unsurprisingly, the jobs most at risk are those that are routine in the sense of employing relatively predictable physical and cognitive tasks. Example sectors are office administration, production, transportation, and food preparation, all of which face high exposure.
Geographically, for the United States the greatest effects will be felt in the heartland. Again unsurprisingly, variations are less across larger jurisdictions than smaller areas, so the difference across states ranges only from 48% of tasks at risk in Indiana and Kentucky to 42% in Massachusetts and New York. Looking in more detail, smaller, more rural areas are more at risk of task substitution, and smaller metropolitan areas are more at risk than larger ones. Yet there are variations even across those categories. Small industrial areas, like Kokomo, Indiana, are as much as 55 percent at risk, while small, especially university, towns, like Charlottesville, Virginia or Ithaca, New York appear relatively insulated. Among larger metro areas, the exposure of less well-educated ones ranges up to 50%, while the areas with as little as 39% exposure are a who’s who of education and technology—San Jose, California, and Washington DC, for example.
Demographics also matters, given that labor markets are already highly segmented by gender and race. Men are more exposed than women, given their predominance in production, transportation, and construction-installation occupations. By contrast, women comprise as much as 70% of the labor force in less exposed occupations like healthcare, personal services, and education. Young people, those aged 16 to 24, are highly exposed given how over-represented they are in food service. Latinx and Blacks are more exposed than whites and Asians, again because of sectors in which they tend to cluster—construction and agriculture for Latinx, and transport for Blacks.
One recent study searched over 16,000 AI patents using a natural language processing algorithm, looking for verb-object connections. Since taking out a patent costs money, those connections may be a better indicator of where investors think AI will make a difference than simple expert interviews. Several of the eight most frequent verbs—“recognize,” “detect,” and “determine”—suggest augmenting human perception. When paired with objects like “face,” “abnormality,” or “illegality,” they point to relatively straightforward measurements at which AI has made rapid progress. Other frequent verbs, like “control,” when connected with objects like “emissions” or “traffic,” suggest relatively routine processes in pursuit of efficiency.
More intriguing were verbs like “determine” or “classify,” which did seem to evoke higher-level human brain work, especially when linked to objects like “relevance” or “data.” “Predict” was even more suggestive when linked to “behavior” or “performance,” suggesting that AI will mimic higher-level human mental processes. Prediction is ubiquitous, all the more so working up the professional ladder in occupations such market research or financial management.
AI vs. Automation, Tasks vs. Jobs
There has been a tendency to lump automation and AI together. The former, and the software that drives it, is in fact likely to hit production jobs, ones that involve routine or rule-based processes, and are lower-paying. A second tendency has been to focus on jobs, rather than tasks. Focusing on tasks, AI, as distinct from robotics, is likely also to affect what we think of as higher-skilled professional occupations. Professionals in, for instance, radiology, optometry, and the law will not get a pass from AI. As one analyst put it: “Whereas low-skill occupations are most exposed to robots, and middle-skill occupations are most exposed to software, it is high-skill occupations that are most exposed to artificial intelligence.”
All that said, the uncertainty about precisely how automation and AI will affect job markets is enormous. After all, most of the jobs that will be created don’t exist now, so we can’t even name them. Amidst all the uncertainty, the optimistic case argues that AI has the distinct possibility to create more jobs than it destroys, a trend which has been consistent with technology advances throughout history. “In the past century, we’ve seen the demise or diminishment of titles like travel agent, switchboard operator, milkman, elevator operator and bowling alley pinsetter. Meanwhile, new titles like app developer, social media director, and data scientist have emerged.”
For the optimists, the efficiency gains from AI will outweigh the transition costs. In addition, AI can enhance the productivity of existing workers, so for certain roles the technology is a great boon. The other side of the argument predicts much more chaos and loss of productivity, at least in the short-term. Harkening back to examples of the past, the Luddite movement was a disruptive response to the ever-increasing mechanization of labor, and so was anger directed at machinery in the 1920s as electric motors were introduced into U.S. factories.
Rarely is anyone either an unabashed optimist or pessimist when it comes to AI–it’s just too unpredictable. The pessimistic view derives from that uncertainty: we can name the tasks and jobs that are likely to disappear but not the new ones as yet uncreated. The optimistic view is rooted in logic and history, asking why this evolution in AI should be different from the groundbreaking technological advances of the past. Of course, the danger with being too optimistic is ignoring the inevitable transition period. After all, it took the country the better part of a century to fully accommodate the industrial revolution.
This time around, too, even if the end-result is a net-positive number of jobs, new technologies will cause at least short-term disruption in the labor markets, putting pressure on governments to do what they can by way of re-training programs, unemployment support, and investment in key industries to mitigate the disruption, even while recognizing that the task will be harder this time around due to a faster pace of job losses—and, perhaps, gains. As one analysis put it: “Throughout history, the long-term benefits of new technologies to average people have been immense and indisputable. But new technologies tend to put people out of work in the short run, and what economists regard as the short run can be many years.”
Beyond jobs, AI will raise existential issues about humans and human cognition. As Henry Kissinger put it: “The Enlightenment sought to submit traditional verities to a liberated, analytic human reason. The internet’s purpose is to ratify knowledge through the accumulation and manipulation of ever-expanding data. Human cognition loses its personal character. Individuals turn into data, and data become regnant.”  We now understand that AI will go far beyond automation, and one example, driverless cars, drives home the dilemmas: if forced by circumstances to choose, how would a driverless care decide whether to run over a child or a grandmother?
There are at least three issues of concern. First, AI may have unintended consequences because it lacks a sense of context: witness the chatbot Tay, released on Twitter by Microsoft in 2016, which was intended to develop friendly conversations with teenagers but had no understanding of “friendly” or “reasonable,” and as a result generated racist, sexist, and inflammatory language. Second, in pursuing its goals, especially winning, AI may press humans to change their thought processes, even their values. AlphaGo beat the world’s Go masters by making strategically unprecedented moves—ones that humans had not conceived. Third, as its algorithms become more complicated, AI may be less and less able to explain how it succeeded. To quote Kissinger again: “What will become of human consciousness if its own explanatory power is surpassed by AI, and societies are no longer able to interpret the world they inhabit in terms that are meaningful to them?… The Enlightenment started with essentially philosophical insights spread by a new technology. Our period is moving in the opposite direction. It has generated a potentially dominating technology in search of a guiding philosophy.”
||Gregory F. Treverton was Chair of the U.S. National Intelligence Council until January 2017. He is now Professor of the Practice at Dornsife College, University of Southern California, Chair of the Global TechnoPolitics Forum, and an SMA Executive Advisor. You can read more of his opinion pieces here.
 Mark Munro, and others, Automation and Artificial Intelligence: How Machines Are Affecting People and Places, (Brookings, January 2019), available at www.brookings.edu/research/automation-and-artificial-intelligence-how-machines-affect-people-and-places/
 Webb, Michael, The Impact of Artificial Intelligence on the Labor Market (November 6, 2019). Available at SSRN: ssrn.com/abstract=3482150 or dx.doi.org/10.2139/ssrn.3482150
 Munro and other, cited above
 Mark Munro and others, What Jobs Are Affected by Ai? Better-Paid, Better-Educated Workers Face the Most Exposure, Brookings Metropolitan Policy Program, November 2019, available at www.brookings.edu/wp-content/uploads/2019/11/2019.11.20_BrookingsMetro_What-jobs-are-affected-by-AI_Report_Muro-Whiton-Maxim.pdf#page=11
 Michael, Webb, “The Impact of Artificial Intelligence on the Labor Market.” SSRN Electronic Journal, 2019, doi:10.2139/ssrn.3482150, p. 46
 See Morgan Frank, and others, “Toward Understanding the Impact of Artificial Intelligence on Labor.” Proceedings of the National Academy of Sciences, vol. 116, no. 14, 2019, pp. 6531–6539, doi:10.1073/pnas.1900949116; and Amit Chowdhry, “Artificial Intelligence to Create 58 Million New Jobs By 2022, Says Report.” Forbes, Forbes Magazine, 18 Sept. 2018, www.forbes.com/sites/amitchowdhry/2018/09/18/artificial-intelligence-to-create-58-million-new-jobs-by-2022-says-report/#52581a6e4d4b
 Insider, “AI and the Future of Work.” Wired, 4 April 2018, available at www.wired.com/wiredinsider/2018/04/ai-future-work/
 “Will AI Destroy More Jobs Than It Creates Over the Next Decade?” The Wall Street Journal, Dow Jones & Company, 1 Apr. 2019, www.wsj.com/articles/will-ai-destroy-more-jobs-than-it-creates-over-the-next-decade-11554156299
 Henry A. Kissinger, “How the Enlightenment Ends: Philosophically, Intellectually—in Every Way—Human Society Is Unprepared for the Rise of Artificial Intelligence, The Atlantic, June 2018, available at https://www.theatlantic.com/magazine/archive/2018/06/henry-kissinger-ai-could-mean-the-end-of-human-history/559124/
 Kissinger, cited above