AI and the jobs apocalypse that wasn’t
The media headlines all tell a consistent story about AI and employment: We’re doomed! ‘Robots may shatter the global economic order within a decade,’ says The Telegraph. ‘A World Without Work’ is coming, The Atlantic predicts. ‘The brave new world of robots and lost jobs’ is upon us, The Washington Post says. Most gloomily, Forbes forecasts, ‘The Robots Will Steal All Our Jobs And No, They Won’t Create New Ones’.
There’s just one thing: All these headlines appeared a decade ago, between 2015-16, yet the sky never fell. In fact, the AI and jobs story has flipped. ‘Collectively,’ the Wall Street Journal sums up, ‘the narrative has shifted from worker-light doomsday scenarios caused by AI to a future in which workers keep their jobs – and get a productivity boost.’
It turns out that ‘Jevons paradox’ and ‘Moravec’s paradox’ were massively underappreciated by the experts.
Jevons paradox is named after a 19th-century English economist who first identified the fact that efficiency improvements for many resources or goods are oftentimes accompanied by an increase, rather than decrease, in total consumption of that resource. This is because lower costs often spur greater demand. This means that AI-enabled automation improvements often encourage all new applications and a boost in overall demand for new systems and services – and the labour that makes them.
Moravec’s paradox is named after a computer scientist who observed in the 1980s that many reasoning and physical tasks that humans find easy are often the hardest to automate, while tasks that humans find intellectually difficult, like complex calculations, are relatively easy for computers. This paradox helps explain why employment remains strong in many sectors in which experts had predicted job dislocations due to AI.
Today, many people are coming around to the simple truth that, as a recent Wall Street Journal headline noted, ‘The Job That AI Was Supposed to Kill Needs More Humans Than Ever.’ While that story was about how court stenographers are actually doing just fine in the AI era, it could have been about many other professions. ‘The threat isn’t that AI can do the job better, legal professionals say. It’s that too few humans are going into the field,’ the article noted. The same applies for the freight business where truck drivers continue to be in hot demand despite years of industry opposition to driverless vehicle technology.
Radiologists are another good example. In 2016, Geoffrey Hinton, the Nobel prize-winning ‘Godfather of AI’, famously said we should stop training radiologists now because AI would take over the field within 5 years. A decade later, the New York Times notes that, ‘The predicted extinction of radiologists provides a telling case study. So far, A.I. is proving to be a powerful medical tool to increase efficiency and magnify human abilities, rather than take anyone’s job.’ Indeed, human radiologists are still in great demand and getting paid over $500,000 on average in the US. AI is simultaneously creating new work opportunities in many other medical fields.
None of this should really be surprising. While technology has certainly disrupted many sectors and professions, on balance, it has always been the engine for greater prosperity, including plenty more jobs. There is no fixed amount of work because human needs and wants are infinite, meaning innovation opportunities are as well. As technology helps solve one problem, it opens up the possibility of addressing others. A 2020 MIT commission on ‘The Work of the Future’ found that, ‘In 2018, 63% of jobs in new occupational titles had not yet been “invented” as of 1940.’ Importantly, ‘[m]any of these new jobs are directly enabled by technology,’ the report noted, and most of those new job titles would have been hard to comprehend just a few generations ago.
Doomsday forecasts persist, however, because the media and academics have powerful incentives to promote tales of a coming techno-apocalypse. ‘Pessimism has always been big box office,’ Matt Ridley noted in his book ‘The Rational Optimist’. That is especially true regarding fears of technologically-induced unemployment. As the Pessimists Archive website notes, ‘Robots Have Been About to Take All the Jobs for 100 Years,’ according to newspaper headline writers.
Meanwhile, there aren’t many downsides for those peddling preposterous predictions of doom. A lot of the Chicken Little headlines from the past were driven by reports from academics and consultancies who whipped up fears to great fanfare, likely knowing they will always get a free pass for misguided fear-based forecasts. In 2015, the technology research firm Gartner Inc. predicted a third of all jobs will be lost to automation within a decade. McKinsey made similar predictions. They were completely wrong, yet both firms are bigger than ever and still cranking out ‘expert’ predictions.
Academics also get a free pass when they peddle pessimistic prognostications. Many of the newspaper headlines mentioned above were inspired by a 2013 report from University of Oxford researchers who warned that 47% of US jobs were at high risk of being automated. Not only were their dire predictions wrong, but most of the professions they thought would lose the most jobs instead witnessed employment growth.
Similarly, in 2016, AI expert Ethan Mollick bet economist Rob Seamans that, within 10 years, robots would take half of all the jobs in the warehouse and storage industry. Mollick was completely wrong. Mollick thought employment would drop to 450,000. Instead, industry employment today is about 1.8 million. Instead of losing half the jobs, employment in the industry doubled.
For many professions, AI is already a multiplier tool that will create new opportunities and jobs well beyond what computers have already provided us. There will be major productivity gains as society benefits from complex machine-human interactions that no one could have predicted. A recent Vanguard report found that, ‘the approximately 100 occupations most exposed to AI automation are actually outperforming the rest of the labour market in terms of job growth and real wage increases.’ A new report from the Federal Reserve Bank of Chicago found that, for the period 2019–2024, exposure to AI ‘does not map mechanically onto job loss, at least in the short run, and that task-based measures of technological exposure are better understood as indicators of occupational restructuring than as direct forecasts of employment decline.’
A final danger with all the false predictions of impending AI-induced doom is that by, ‘preparing for the wrong labour-market shock,’ economist Stephen Lewarne notes, the government ‘enacts policies that make it more difficult to adapt’. It seems like every other pundit and policymaker these days has a proverbial Big Plan for solving future labour market problems that probably will never come about. They have no idea what the future holds, but their worst-case thinking will inspire costly, counter-productive proposals that will result in ineffective, poorly targeted interventions that will likely only prevent the transition to a better future for workers and the public.
Humility and prudence are virtues in very short supply in these discussions. The best thing for policymakers to do today is clear away barriers to economic dynamism and new opportunities for workers, creating space for technology to be the amazing job-creating engine it long has been.