Daron Acemoglu Won a Nobel for Asking ‘Progress for Whom?’

When people talk about Acemoglu Power and Progress, they usually reach for the Nobel first — and fair enough. In 2024, Daron Acemoglu, an economist at MIT, shared the Nobel Memorial Prize in Economic Sciences with Simon Johnson and James Robinson, honoured for decades of work on why some societies grow rich and free while others stay poor and unfree. But if you want to understand what Acemoglu is actually for, the Nobel is the wrong starting point. The right one is a single, stubborn question that runs underneath everything he has written: progress, yes — but progress for whom?

That question sounds almost naive until you sit with it. We are trained to treat technological progress as a rising tide. Build a better machine, invent a faster process, automate a task, and prosperity supposedly spreads outward on its own. Acemoglu's life's work is a patient, evidence-heavy argument that this is not how history actually worked — and not how the future will work either, unless we make it. The gains from a new technology are not distributed by physics. They are distributed by choices, institutions, and power.

Who Daron Acemoglu is, and the question under his work

Acemoglu is one of the most cited economists alive, an institution unto himself at MIT, and he writes across an unusually wide range — political economy, labour markets, growth theory, the deep history of prosperity. With his frequent collaborator James Robinson, he wrote Why Nations Fail, which argued that the difference between rich and poor countries comes down not to geography or culture but to institutions: whether a society's rules are inclusive, spreading opportunity and protecting broad participation, or extractive, funnelling wealth and power to a narrow elite. It's a book about who gets to hold the levers, and what they do once they have them.

Hold that framing in your head, because it never really leaves his work. The through-question — who holds power, who benefits, who pays — simply migrates from the study of nations to the study of machines. And that migration is the book that matters most for our moment: Power and Progress, written with Simon Johnson, an economist and his fellow 2024 laureate.

The core argument of Power and Progress

The central claim is deceptively simple and quietly radical. Technological progress does not automatically lift everyone. Whether a new technology broadly improves people's lives or mainly enriches a small group who own and direct it is not written into the technology itself. It depends on choices — how the tool is designed, what it's pointed at, who controls it, and what countervailing forces (unions, regulators, competitors, an organised public) exist to bend the gains outward.

Acemoglu and Johnson draw a sharp distinction that is worth carrying around with you. A technology can be used to augment workers — making people more capable, creating new tasks, raising what human labour is worth. Or it can be used to automate — replacing workers, cutting them out of the loop, concentrating the returns among those who own the machines. Both are "progress" in the narrow sense of doing more with less. But they land on very different people. Augmentation tends to share the surplus; pure automation tends to hoover it upward.

The gains from a new technology are not distributed by physics. They are distributed by choices, institutions, and power.

This is the argument that, to me, sits right at the centre of how how technology gets captured. A tool arrives full of genuine promise. The promise is real. But the direction it gets pushed — augment or automate, share or concentrate — is decided by whoever has the power to decide, and they rarely decide against themselves. Progress and capture are not opposites. Capture is what happens to progress when nobody with an interest in sharing it is strong enough to insist.

The history he marshals: the long, grim first act

Acemoglu's favourite piece of evidence is one most of us were taught to skip past. We tell the Industrial Revolution as a triumph — steam, factories, the great escape from poverty. And in the long run, it was. But the long run hid a brutal opening act. For roughly the first several decades of British industrialisation, the people doing the work did not obviously win. Hours were savage, cities were filthy and dangerous, child labour was routine, and by several measures the living standards and even the height and health of ordinary workers stagnated or declined while output and profits climbed. The machines were productive. The productivity did not reach the workers' tables for a long, painful time.

What changed it, in his telling, was not the technology maturing on its own. It was power shifting. Workers organised. The franchise widened. Unions won bargaining strength, reformers won factory laws and public health, and eventually a share of the gains was forced outward through political struggle rather than granted by the machines. The lesson Acemoglu draws is not that industrial technology was bad. It's that the broad-based prosperity we retroactively credit to "the machines" was actually the product of a fight over who the machines would serve — a fight that took the better part of a century and was not guaranteed to be won.

He reaches further back, too — to medieval agricultural improvements whose surplus flowed to landowners and the church rather than the peasants who worked the land. The pattern rhymes across centuries. Someone invents a better way to produce. The output rises. And then a separate, political question decides who eats the difference. Sometimes the answer is "almost everyone." Often, for a long time, the answer is "the people who already held power."

What he says about AI

All of this is prologue to the argument he most wants us to hear, because we are living inside the next iteration of it. Artificial intelligence, in Acemoglu and Johnson's reading, is not a fixed thing with a fixed effect. It is a direction we are still choosing. It could be built to augment — to give a nurse better diagnostic support, a teacher more time with students, a tradesperson sharper tools, an ordinary worker capabilities they didn't have before. Or it could be built to automate — to strip tasks out from under people, cut headcount, and route the returns to a handful of firms that own the models and the compute.

Nothing about the underlying science forces one path over the other. What tilts the odds is who is making the decisions and what they're optimising for. And right now, Acemoglu warns, the incentives lean hard toward automation and concentration: a business culture that treats cutting labour costs as the obvious win, and a technology sector where control over the models, the data, and the infrastructure sits with very few hands. This is exactly the terrain of will AI take my job — and his answer is bracingly unromantic. AI won't "take" your job as a matter of destiny. People will choose to deploy it in ways that take your job, if that's the cheaper path and nothing stops them.

AI won't take your job as a matter of destiny. People will choose to deploy it in ways that take your job — if that's the cheaper path and nothing stops them.

The corollary is the hopeful half, and Acemoglu insists on it: if the direction is a choice, it can be chosen differently. Redirect the research toward tools that make workers more valuable rather than redundant. Change the tax treatment that quietly subsidises replacing people with machines over hiring them. Build worker voice and public oversight into how these systems get deployed. Break up or discipline the concentration of control. None of this is a fantasy of stopping technology — it's a demand that we steer it, the way the nineteenth century eventually, painfully learned to steer the factory.

Where critics push back

It would be dishonest to present all this as settled. Acemoglu is influential precisely because he makes strong, falsifiable claims — and strong claims attract sharp disagreement. Some economists argue he leans too hard on institutions and underweights other drivers of growth like geography, human capital, or plain accumulated knowledge. Critics of Why Nations Fail have long said the inclusive/extractive dichotomy is cleaner on the page than in the messy record of real countries, and that the causal arrows are harder to pin down than the framework implies.

On the technology side, more optimistic economists contend that automation has, historically, kept generating new jobs and new tasks faster than it destroyed old ones, so the "this time is different" worry about AI may be overstated. Others question some of the empirical estimates about how much automation has already suppressed wages, or push back on his more recent, deliberately modest forecasts about AI's near-term productivity impact — arguing he is too pessimistic in one direction or another. These are real debates, not knockout blows, and Acemoglu engages them. But anyone citing him should be honest that the augment-versus-automate story is a powerful lens, not a proven law.

Why this maps onto who captures the gains

Strip away the citations and what Acemoglu keeps circling is the question I find myself returning to again and again: not does technology create wealth, but who captures it. That's the same question whether you're looking at a steam loom in 1815, a search engine in 2005, or a large language model today. The value is real. The fight is over the flow — where it pools, who guards the pool, and whether anyone with the power to redirect it has a reason to.

This is why his framework reads as a natural companion to the debate over who owns AI. Ownership of the models, the compute, and the data is not a technical footnote; in Acemoglu's terms it is the whole game, because ownership is where the power to choose augment-or-automate actually lives. Concentrate that ownership and you tilt the whole system toward concentration of the gains — almost regardless of what the engineers intend. It's also why his work speaks to the darker forecasts around technofeudalism: a world where a few platform owners collect rent on infrastructure everyone else is forced to use is, in his vocabulary, just an extractive institution wearing a hoodie. The names change. The pattern — value created broadly, captured narrowly — is old.

What I admire about Acemoglu is that he refuses both of the easy stories. He won't tell you technology is a benevolent force that fixes itself, and he won't tell you it's a doom you're powerless against. He tells you something harder and more useful: the outcome is contested, it has always been contested, and the side that organises and insists tends to be the side that gets a share. The early Industrial Revolution didn't become broadly prosperous because the machines got kinder. It became prosperous because people fought, for decades, over who the machines would serve — and eventually enough of them won.

That is the real weight of the question that earned him a Nobel. "Progress for whom?" is not a rhetorical flourish. It is an instruction. It tells you to look past the demo and the stock price and ask where the value is landing, who's guarding it, and what would have to shift for it to land somewhere else. With AI, that fight is happening right now, mostly out of sight, in decisions about how the tools get designed and who gets to own them. Acemoglu's contribution is to make sure we can't pretend it isn't a fight — or pretend that the outcome was ever going to take care of itself.