Nobel laureate in economics names 3 things in AI to watch

AI Agents: The Threat Is Exaggerated, But Worth Watching

One of the major technical breakthroughs of recent years has been AI agents—systems capable not just of answering questions, but of independently performing tasks. Companies are actively positioning them as replacements for entire teams of employees.

Daron Acemoglu disagrees with this view. In his opinion, agents are more likely to assist with specific fragments of work rather than replace humans entirely. As an example, he cites a radiologic technologist, who has around 30 different tasks—ranging from recording patient history to archiving images.

Humans easily switch between formats and systems, whereas an agent would require a separate tool or protocol for each task. Until agents learn to transition between tasks as freely as humans do, most professions face no threat.

Economists on the Payroll of Big Tech

The second trend that concerns Acemoglu is the mass hiring of economists by major AI companies. OpenAI recruited Ronnie Chatterji from Duke University as its chief economist, as well as Jason Furman—a Harvard professor and former advisor to Barack Obama—for joint research on the impact of AI on employment. Anthropic assembled an advisory board of 10 leading economists. Google DeepMind hired Alex Imas from the University of Chicago as its Director of AGI Economics.

Daron Acemoglu understands the logic: public skepticism regarding AI is growing, and companies need to shape the narrative around their technologies. However, he openly worries about another issue: an increasing share of the most influential research on the impact of AI on the labor market will come from those who benefit from positive conclusions.

AI Applications: The Technology Exists, Convenience Does Not

The third point concerns accessibility. Acemoglu suggests comparing AI to Word and PowerPoint—programs that revolutionized office work at the time. Anyone could install them and start using them immediately. With AI, things are different: even if you can communicate with it in natural language, deriving real benefit from it in the workplace is far from immediate.

This is precisely why AI has not yet had a noticeable impact on either labor productivity or employment. Daron Acemoglu is watching to see whether applications will emerge that make the technology as simple and understandable as office programs were in the 1990s.

Uncertainty as the Main Takeaway

Acemoglu acknowledges that the data currently favors his view—studies repeatedly show that AI is not affecting employment levels or layoffs. However, rhetoric is outpacing reality. Talk of an AI apocalypse is heard at Bernie Sanders rallies and in grocery store lines, and some economists, previously skeptical, are beginning to admit that something truly large-scale might happen.

At the same time, he warns that in the near future, data on AI will contradict each other. On one hand, there are stories about college graduates finding it increasingly difficult to secure jobs. On the other hand, there is no measurable impact of AI on economic productivity.

According to Acemoglu, there is currently a vast amount of uncertainty. And this is the most telling aspect of the AI economy right now: complete certainty in the rhetoric, and complete uncertainty in everything else.