Into The Impossible: Daron Acemoglu on AI Hype, Jobs, and Liberal Democracy

SPEAKER_1 (0:01)

When science, power, and technology collide, the real question is not just what can be built, but who gets to decide.

Brian Keating sat down with Daron Acemoglu of MIT, a Nobel Prize-winning economist whose new book tackles liberal democracy's role in scientific progress, social media, and AI. And Acemoglu doesn't hold back. He says liberal democracy offers "a much better foundation for science" because it celebrates inquiry, experimentation, and fallibility. Science advances by challenging settled authority, not by bowing to it. That brings us to China.

Acemoglu admits he wouldn't be comfortable there because of surveillance and speech restrictions, yet he understands why researchers might be tempted by better funding. And here's the thing—China isn't just copying us. They're investing heavily in AI, robotics, manufacturing, and engineering talent. Right, but history complicates this picture. Take Galileo, who worked under an illiberal Catholic Church and still helped science flourish, though Acemoglu notes there were always limits, delays, and bans. The lesson? Authoritarian systems never close every door, but they make inquiry harder. What about academia itself?

Acemoglu makes an interesting distinction between democratic politics and expert judgment.

He argues peer review and grant decisions aren't illiberal just because they're not majoritarian. The challenge is keeping them open and willing to question received wisdom. As he puts it, "consensus is no guarantee of truth."

He actually says heresy has a place, especially in academia, where useful dissent can be rewarded if it's serious and defensible. But outside universities? Much harder to be "heretical at Google or Microsoft," where conformity gets baked into corporate culture. Then there's social media. Acemoglu says early hopes that platforms would democratize journalism were wrong because companies optimized for engagement, outrage, and echo chambers. His answer? Regulation, including one blunt idea: eliminate algorithmic feeds. That concern extends to power and inequality. He argues modern society lets extreme fortunes translate into political influence without guardrails. And the tax system distorts incentives by treating income from ownership differently from labor. On AI, he pushes back against the hype. Even with newer agentic systems, he thinks productivity gains would be "an order of magnitude less than what the industry claims." His bigger worry is jobs. If AI only aims at AGI-style automation, it could deepen inequality. Instead, he wants AI to create new tasks and capabilities. And he draws a crucial line between general models and domain-specific tools. Systems like AlphaFold, built for one job, are easier to align with human goals than models trained on everything.

Looking ahead, Acemoglu fears liberal institutions may weaken over the next twenty years. He argues scientists have a duty to be less hubristic and more modest. His closing message? "Fallibility should affect not just our practice of science" but also how scientists communicate with wider society.