Tech • AI • Robotics • Game
Will AI cause utopia or disaster? | The Economist
6/10TL;DR
Despite deep public anxiety about artificial intelligence, current labour-market data in the OECD and America do not show an AI-driven jobs collapse, even as the technology could still reshape work, growth and public policy.
KEY POINTS
Employment remains strong
Public concern about job security has climbed sharply, with one international survey across 19 countries finding that 40% of workers fear losing their jobs within five years because of AI and automation, while 60% worry about machines taking over roles. Yet employment data point in the opposite direction for now. The share of the working-age population with a job in the OECD is at a record high, and the number of white-collar jobs in America has risen by about 3 million since ChatGPT was launched in late 2022.
ChatGPT is an imperfect benchmark
Economists often treat the arrival of ChatGPT as a useful starting point for measuring AI’s effect on labour markets, but that shortcut can distort the picture. Early generative AI tools were still limited, making it unlikely that any immediate slowdown in hiring was mainly caused by AI. Weakness in some entry-level white-collar recruitment appears to have started earlier, after firms overhired during the pandemic and then pulled back as interest rates rose in 2022.
Pandemic hiring and cost cutting matter more so far
During the pandemic, heavy demand for digital services encouraged companies to expand payrolls aggressively. When borrowing costs increased, that hiring boom faded and many employers looked for savings through outsourcing and tighter recruitment. Those forces offer a more convincing explanation for recent softness in parts of the office-job market than a sudden wave of AI displacement.
History offers little support for an immediate jobs apocalypse
Fears of mass technological unemployment are not new. Computers once prompted warnings that office workforces would be slashed, yet the labour market adapted instead of collapsing. One estimate suggests that roughly half of America’s employment growth between 1980 and 2010 came from entirely new occupations, underscoring a long pattern in which innovation destroys some tasks but also creates fresh categories of work.
Warning signs would likely appear first in productivity and wages
If AI were truly beginning to eliminate jobs at scale, economists would expect clearer macroeconomic evidence. The likeliest early signals would be sharply faster productivity growth paired with weak real-wage growth, especially in America, still regarded as the world’s frontier economy. A recession could also accelerate labour-market change, because downturns tend to eliminate less productive roles and speed corporate restructuring.
The upside case is unusually large
Optimists argue that advanced AI could deliver an extraordinary boost to economic growth by accelerating discovery and carrying out a rising share of productive tasks. One bullish projection suggests annual growth could exceed 20% once AI can perform 30% of tasks. In healthcare, researchers at MIT are using AI in work targeting antibiotic-resistant infections such as MRSA, highlighting the technology’s potential to improve outcomes beyond the office.
Education may be one of the earliest broad gains
AI could also widen access to high-quality learning by giving children tools once available mainly to wealthy families, including personalised tutoring and tailored materials. In Nigeria, pupils in a six-week after-school programme used a chatbot twice a week and improved their English scores by the equivalent of nearly two extra years of schooling. Such results suggest AI may lift human capital even before it transforms the wider economy.
Long-term risks remain serious
More pessimistic assessments focus less on near-term job losses and more on systemic danger. Some researchers place the chance of an AI-driven catastrophe by 2100 at 5%, and judge major harm within the next 25 years to be more likely than not. Concerns include systems relentlessly pursuing poorly specified goals, consuming critical resources, or improving themselves by rewriting code without meaningful human oversight.
Policy could face pressure even without total automation
Even a less extreme future could strain existing tax and welfare models if labour’s share of income falls or if fewer people earn traditional wages. That possibility has helped drive calls in Silicon Valley for stronger state intervention and, in some scenarios, a universal basic income. Still, a middle path remains the most likely outcome, with human interaction preserving demand for many roles even as tasks within them change.
CONCLUSION
The evidence so far does not support the idea that AI has already triggered mass unemployment. The bigger question is how quickly productivity, wages, education and public policy begin to shift as the technology becomes more capable.
Ask a question