Acemoglu challenges AI boom forecasts and warns of job risks
The MIT Nobel laureate told Fortune that AI’s productivity gains may be modest and that mass graduate joblessness could strain democracy.
By Hana Yoshida · Markets Reporter
3 min read
MIT economist Daron Acemoglu is pushing back against Wall Street’s bullish assumptions about artificial intelligence, arguing that the technology is likely to deliver only modest productivity gains over the next decade. In an interview with Fortune, the 2024 Nobel Memorial Prize winner said the debate over AI also misses the political danger posed by concentrated corporate power and job displacement.
Acemoglu, whose work has focused on institutions and prosperity, estimated that AI will add about 0.55% to total factor productivity over 10 years. He told Fortune that only about 5% of tasks are likely to be profitably automated in the near term, which he said would translate into roughly a 1% to 1.5% increase in GDP.
He also gave a blunt assessment of the public debate around AI and capitalism. Asked how much of the AI discussion he regards as intellectually serious, Acemoglu put the figure at about 20%, telling Fortune that roughly 80% is “brainless,” while clarifying that he meant speculative or fictional rather than stupid.
Markets, power and institutions
Acemoglu told Fortune he dislikes the word “capitalism” because it treats very different economies as if they operate under one shared model. He cited Sweden, Egypt, Argentina, Honduras, the United States, South Korea and Japan as examples of countries grouped under the same label despite being organized in sharply different ways.
His preferred distinction, developed with James Robinson in books including Why Nations Fail and The Narrow Corridor, is between inclusive and extractive institutions. Fortune reported that Acemoglu applies that frame to AI by asking whether the technology broadens opportunity or concentrates wealth and control.
Acemoglu argued to Fortune that today’s leading AI companies resemble extractive institutions because of concentrated ownership, regulatory influence and business models built around data and attention. He said the focus should be on monopoly power, inequality and the displacement effects of AI rather than broad arguments about capitalism as a single system.
Why he doubts sweeping productivity gains
Fortune reported that Acemoglu’s skepticism rests on how automation creates economic value. He said machines raise productivity only when they perform tasks much more cheaply or effectively than people, and that marginal improvements can be offset by the cost of integrating new systems.
Acemoglu said the more promising path would be AI that helps workers perform new tasks, rather than speed up existing ones. He cited podcasts as an example of a format that expanded demand for news, telling Fortune that AI would need to create similarly new services or capabilities to deliver stronger gains.
He also questioned research that finds large AI productivity benefits, saying much of it focuses on narrow, clearly defined tasks. According to Fortune, Acemoglu said those studies do not reflect much of the wider economy, where work is often complex and context-dependent.
For very large productivity gains, Acemoglu told Fortune, AI systems would need to approach artificial general intelligence. He said current models fall short in areas that matter in professional settings, including reading social context, making cross-domain connections and handling work where human judgment is central.
Concerns over graduates and democracy
Acemoglu said large-scale job losses among young graduates would pose risks beyond the labor market. If 30% or 40% of new university graduates were unable to find work, he told Fortune, the result could threaten democracy and social peace.
He said past episodes of widespread youth frustration have been associated with revolutions, while adding that outcomes are unpredictable. Acemoglu also noted that social media platforms such as Instagram, TikTok and Twitter introduce a factor that earlier generations did not have.
Acemoglu called for a broader public discussion about what society wants from AI, with attention to wages, jobs, shared prosperity and dignified work. He also told Fortune that global AI governance should include U.S.-China cooperation on areas such as disease control, productivity, safety rules and useful applications, though he said current politics make that difficult.
This story draws on original reporting from Fortune.