Nobel-winning economist reveals 3 AI trends to watch
▼ Summary
– Public anxiety about AI-driven job loss persists, but data continues to show no significant effect on employment rates or layoffs, supporting economist Acemoglu’s cautious view.
– Agentic AI, which can operate independently, is being marketed as a replacement for human workers, but Acemoglu argues it is better suited as a tool for specific tasks rather than entire jobs.
– Whether AI agents will disrupt employment depends on their ability to seamlessly switch between the many varied tasks humans perform naturally in a single job.
– AI companies like OpenAI, Anthropic, and Google DeepMind are hiring in-house economists to research the impact of AI on jobs, signaling growing concern about the issue.
– Acemoglu’s core thesis remains unchanged, but he now focuses on the limitations of AI agents in handling task orchestration rather than on fears of imminent AGI.
Two years on, the measured perspective of Nobel laureate Daron Acemoglu has yet to gain widespread traction. Fears of an AI-driven jobs apocalypse now surface everywhere, from Senator Bernie Sanders’ campaign speeches to casual conversations overheard at the grocery store. Even some economists who were once skeptical are warming to the idea that a major disruption could be on the horizon. Just last week, a California gubernatorial candidate proposed taxing corporate AI usage to compensate those affected by “AI-driven layoffs.”
On the surface, the data still supports Acemoglu’s cautious stance. Numerous studies continue to show that AI has not significantly altered employment rates or triggered mass layoffs. Yet the technology itself has evolved dramatically since his earlier predictions. I sat down with Acemoglu to explore whether recent breakthroughs have shifted his core thesis and to learn what actually keeps him up at night if an imminent AGI doesn’t.
AI agents represent one of the most significant technical advances since Acemoglu’s paper. These tools go beyond simple chatbots, operating autonomously to complete user-defined goals. Because they can work independently rather than just answering queries, companies are increasingly marketing them as a one-to-many replacement for human employees.
“I think that’s just a losing proposition,” Acemoglu says. He argues that agents should be viewed as tools to augment specific tasks within a role, not as flexible enough to take over an entire job.
His reasoning stems from research he began in 2018 on the complexity of work. Consider an x-ray technician, who juggles roughly 30 distinct tasks, from recording patient histories to organizing mammogram image archives. A human can seamlessly shift between formats, databases, and workflows. But how many separate tools or protocols would an AI agent need to replicate that fluidity? The real test of whether agents will amplify AI’s impact on employment hinges on whether they can eventually master the task orchestration that humans perform naturally. AI companies are racing to prove their agents can operate independently for longer stretches without errors, sometimes overstating their results. Yet Acemoglu believes many jobs will remain safe if agents cannot smoothly transition between tasks.
The new hiring spree also caught his attention. For years, Big Tech has offered astronomical salaries to recruit AI researchers. But I asked Acemoglu about a different trend I’ve observed: AI companies are quietly building in-house economics teams. OpenAI hired Ronnie Chatterji from Duke University in 2024 as its chief economist, and last year announced that Chatterji would collaborate with Harvard economist Jason Furman, a former advisor to Barack Obama, to study AI’s impact on jobs. Anthropic has assembled a group of 10 leading economists for similar research. And just last week, Google DeepMind announced it had hired Alex Imas, an economist from the University of Chicago, as its “director of AGI economics.”
(Source: MIT Technology Review)
