283 | Daron Acemoglu on Technology, Inequality, and Power
What this covers
Daron Acemoglu, an MIT economist and author of Power and Progress, joins Sean Carroll to examine whether technological progress automatically improves the lives of ordinary workers. The conversation centers on Acemoglu's core claim: technology itself is neutral, a reflection of collective knowledge that can be deployed many ways, and the distribution of its benefits depends entirely on institutions, political power, and deliberate choices about which technological paths to pursue. To ground this argument historically, Acemoglu traces the British Industrial Revolution, where for roughly ninety years productivity soared while workers' real incomes stagnated, wages collapsed in mechanized sectors, and life expectancy in manufacturing cities plummeted. Improvement came only through contentious institutional transformation—the shift to democracy, the legalization of trade unions, and movements like the Chartists—alongside a reorientation of technology itself toward raising worker productivity rather than eliminating workers altogether.
The discussion ranges across multiple domains to show how inclusive versus extractive institutions shape technological outcomes. Acemoglu examines medieval peasants forced to use elite-controlled windmills at imposed prices, the American Progressive Era's successful assault on concentrated corporate power, and Sweden's transformation from deep inequality into shared prosperity through corporatist negotiation between business, labor, and government. He introduces the concept of power to persuade—ideological soft power unrelated to coercion—arguing that tech industry dominance operates through inspiration and narrative rather than force, and that concentrated soft power in media representation of tech leaders erodes even nominally inclusive systems. On artificial intelligence specifically, he argues that large language models need not eliminate workers if deployed as contextual tools raising marginal productivity rather than as replacements. Solutions include policy interventions like digital ad taxes and antitrust action, plus the restoration of countervailing institutions—unions, regulators, and civil society seats at technological steering tables.
Acemoglu argues that technological progress does not automatically benefit ordinary workers; whether it produces shared prosperity or extraction by elites depends on deliberate choices about institutions, the direction of technology, and the distribution of power, including soft 'power to persuade.'
- The first ~90 years of the British Industrial Revolution made workers worse off despite rising productivity, and improvement only came through contentious political and institutional change.
- Technology is not deterministic: the same knowledge can be used to automate away workers or to raise their marginal productivity, and which path is taken is a social choice.
- Steering technology toward shared prosperity requires countervailing powers — unions, regulators, civil society — plus specific policies like digital ad taxes and antitrust.
Technological progress does not automatically improve worker welfare; power and institutions determine who captures gains from productivity.
- During the first roughly 90 years of the British Industrial Revolution (c.1750 to the late 1840s), real incomes of the average worker did not increase much even though productivity rose and some made fortunes; in the most dynamic sectors workers fared badly, with weaver wages falling to about 30% of their pre-mechanization level, alongside horrible working conditions and life expectancy in cities like Manchester falling to around 30 years at birth.
“Real incomes of the average worker did not increase much during this 90 year period, even though productivity rose and some people made fabulous fortunes, the real wages of the workers in the most dynamic sectors of the economy did really badly.”
- The eventual improvement in workers' conditions after the 1850s was not automatic but required an enormous amount of contentious political change—the UK shifting from oligarchy to democracy through adult suffrage, the legalization and recognition of trade unions, and movements like the Chartists (who collected 3 million signatures in the 1840s yet had their leaders jailed)—plus a shift in how technology was directed toward making workers more productive rather than merely automating and exploiting them.
“It didn't happen by itself. It was a very, very contentious process. First of all, it took an enormous amount of political change.”
- Medieval technological improvements like crop rotation and windmills were genuinely innovative and more efficient, but because society was politically and socially unequal, elites forced peasants to use the elite-controlled windmills at imposed prices and banned alternatives, so the elite captured all the surplus and the peasants experienced no gain—spending the proceeds on luxurious consumption such as ever-bigger monuments.
“the elite said, no, no, we're not gonna allow that. We're gonna say this is the better technology. You have to come and use it, and you cannot use an alternative technology. And you're gonna do it at the prices that we impose on you.”
- Institutions normally protect ordinary people from the worst abuses, but when a powerful new technology arrives it can give elites a better way to extract wealth from everyone else, and only later can institutions catch up to spread the benefits more broadly.
“But then a new technology comes along, and it sort of gives the elites a better way to extract from everybody else until the institutions can catch up later.”
- Even if technology raises a worker's contribution to output, workers may capture none of the gains when laws and institutions are stacked against them and employers are organized to repress them—as under the UK's Masters and Servants Acts, where a worker who tried to quit for a better job could be taken to court and jailed—so shared prosperity requires both productivity-enhancing technology and a balanced distribution of power embedded in institutions and social norms.
“In the UK, there were all these masters and servants acts. So if you told your employer, "I have a better job offer, I wanna quit." They could take you to court and put you in jail.”
- The 'factory of the future' parable—a factory with only a man and a dog, where the man feeds the dog and the dog keeps the man from touching the equipment—illustrates that even if such a factory doubles its output, there is no reason to pay or hire more workers because their marginal productivity (contribution to output) is near zero, even though average productivity is huge; this is the danger if technology is used only to eliminate workers.
“the factory of the future will have two employees, a man and a dog, the man is there to feed the dog and the dog is there to make sure the man doesn't touch the equipment.”
Soft power—ideological influence through media, networks, and presence—is the dominant form of power today and operates through social psychology.
- Power is neither purely coercive (Marxist guns-and-troops view) nor irrelevant to economics; the crucial form today is non-military 'power to persuade' or ideological/soft power, which is poorly understood but operates through social psychology (what makes ideas palatable), social networks that propagate and lend prestige to ideas, and who is at the table—so the tech industry's power, having nothing to do with tanks and guns, comes from its power to inspire, persuade, and present its agenda as a fait accompli that the rest of society accepts.
“In the book, we call that power to persuade. And I think if you want to understand power to persuade, you need to think about social psychology.”
- Even where economic and political institutions are nominally inclusive, if soft power becomes concentrated in a very small group it will start bearing down on and eroding any inclusive system—a danger Acemoglu sees in the United States, which has been broadly democratic and opportunity-creating but where the concentration of soft power now creates real tension for inclusive economic and political institutions.
“if soft power is very much in the hands of a very small group, that's going to start bearing down any inclusive system that you have.”
- Media plays a central role in soft power: continuously talking about tech leaders the way we talk about movie stars elevates them and increases their power to persuade, whereas if trade union leaders were on TV every day we would think about workers' conditions very differently—so the power to persuade interacts strongly with which voices the media platforms.
“In the book, we call that power to persuade. And I think if you want to understand power to persuade, you need to think about social psychology.”
- Elites engaging in extraction are not Hollywood-style evil; they genuinely convince themselves they are working for the social good, and historically extraction has consistently been couched in the name of the common good—from keeping peasants as peasants, to the UK enclosures movement, to Soviet collectivization—which should make one skeptical when tech companies claim they are disrupting an industry for the common good.
“they're not evil in the sense of Hollywood movie evil, everybody convinces themselves that they're doing something that's good.”
- The big tech companies today exert control over so many aspects of the economy and society that their dominance exceeds that of Standard Oil—often imagined as the octopus with tentacles everywhere—and would have made Rockefeller jealous.
“standard oil was nothing relative to the... Compared to the big tech companies today. Their control over so many aspects of the economy and society would have definitely made Rockefeller jealous.”
- Acemoglu's thinking has shifted since Why Nations Fail was completed in 2010: where the book was broadly optimistic that US institutional flexibility could overcome its challenges and cautiously optimistic about the Arab Spring, he would today be less optimistic about US institutions, partly because of the tech sector's dominance—which was not easy to see in 2010—and the erosion of institutional guardrails.
“we were broadly optimistic that US institutions and institutional flexibility would be able to overcome some of these challenges. Well, today, I would be less optimistic about that.”
Technology is a choice reflecting collective values, not a predetermined path; digital and AI knowledge can eliminate or empower workers depending on direction chosen.
- Science is an evolutionary, decentralized process that cannot be centrally controlled from the top—attempts to centralize it fail—but because there are far more things to work on than can be pursued, what scientists focus on is a social decision shaped by career concerns, monetary incentives, and business models, and that decision may not be the socially optimal one, so government can legitimately set agendas and steer with a hands-off approach (as in the Manhattan Project, aerospace, and antibiotics) without dictating how research is done.
“Science is an evolutionary process... Whenever you try to centralize science, that's not gonna work... at any point in time what we focus on is a social decision.”
- By 'vision' is meant the aspirations of powerful people and society about what is acceptable and desirable; if companies and society treat rushing toward AGI as the most exciting and acceptable goal even at the cost of safety and inequality, that is what we get, whereas if different actors with different visions—such as trade unions seeking tools to help their members—were at the table, outcomes would differ, and nothing in the market system automatically selects the right vision because markets are good at allocating goods but not at choosing between technological paradigms.
“if we have a bunch of companies who think that the most exciting thing is to try to rush towards AGI, and that's also socially acceptable and desirable... That's what we're gonna get, because it's the vision of these powerful actors.”
- Technology is just a reflection of our collective knowledge that can be applied in many different ways; there is no predetermined path of technological progress, as illustrated by nuclear physics, which could be used to build reactors for energy or to make weapons—so with digital and AI knowledge we likewise have a choice between eliminating workers or making them more productive.
“you can use them for building nuclear reactors for energy, or you can use them for making nuclear weapons, very different implications.”
- If the tech sector prioritized increasing worker productivity and providing better tools for workers—done collaboratively rather than in a way that tries to expropriate firms—it would ultimately be good for firms too, because more productive workers make firms more innovative, able to provide higher-quality services and expand into new areas, mirroring the non-adversarial Swedish bargain.
“if we can get into a different path where the tech sector prioritizes increasing worker productivity, providing better tools for workers, that's actually gonna be ultimately good for firms.”
- Recognizing two things changes the whole debate: it is both socially desirable and technically feasible to have a different direction for AI—desirable because using AI to help workers, improve information, and aid the disadvantaged is much better than the current trajectory, and feasible because it is not a crazy dream, as shown by Taiwan's use of AI to build new democratic platforms, hold officials accountable, and create new public squares—so claims that vast funding for an automating AGI path is the only route to a better future can be rejected.
“it is socially desirable, and B, it's technically feasible to have a different direction of AI.”
Historical precedent shows contentious political change and democratic pressure successfully shift institutions and redirect technology toward worker benefit.
- The US Progressive Era reforms emerged from conditions that looked completely hostile to change—senators not directly elected and seen as captured by individual millionaires, an unfree media, no antitrust regulation, no federal income tax, and no consumer-protection agencies—yet once a movement coalesced around middle-class life, controlling big corporations, and reform, it became self-sustaining and transformed US society toward more equal distribution and constrained corporate power, demonstrating that non-revolutionary change once thought impossible can happen.
“Senators were not directly elected and the general perception was that each senator was in the pocket of one billionaire, or millionaire at the time... there wasn't proper antitrust regulation, there wasn't a federal income tax.”
- Renewable energy's rise to cost-competitiveness was driven by policy, not unaided markets: regulations (including in California), Germany's pioneering legislation requiring investment in solar panels, innovation subsidies, and modest carbon taxes redirected energy research, after which Chinese firms scaled solar manufacturing and exploited learning-by-doing, cutting thinner wafers and improving panels until the cost of solar-generated energy dropped more than tenfold in about 15 years—showing government can steer innovation while respecting science's decentralized nature.
“We got there not just because the market process suddenly decided by itself to allocate more resources; it's because of policy.”
- Sweden was historically a very unequal economy with deep capital-labor conflict, not a natural cooperative paradise; the Great Depression and the 1932 rise of Social Democrats—who had severed links with Marxism about 20 years earlier and committed to democracy and reform—produced a corporatist model where business, government, and labor unions negotiated agreements good for workers but not so adversarial as to deter business investment, with the result that businesses thrived and remained innovative while wages grew rapidly and both wealth and income inequality declined substantially.
“the Swedish Social Democratic Party, or the Workers Party was very, very early, about 20 years before then completely had severed its links with Marxism. They were committed to democracy and reform”
- The far-left claim that capitalism and the market economy are always doomed to be horrible for workers is rejected on the basis of history: the second half of the 19th century was far better for workers than the first half, and in the three-plus decades after World War II across the industrialized world workers' wages increased faster than GDP, producing genuine shared prosperity.
“If you look at what happened throughout the industrialized world and three, three and a half decades that followed World War II, workers' wages increased faster than GDP. And there was a real sort of flavor of shared prosperity.”
- Steering technology in a beneficial direction requires countervailing powers, not just agreement among commentators: just as the countervailing powers against car companies came from trade unions (limiting wages and conditions) and regulators (forcing better brakes, seat belts, and tires that companies resisted), today we need civil society, a better-functioning democratic process, and worker organizations at the table to counter concentrated tech interests.
“You really need countervailing powers. Where did the countervailing powers against car companies came... it came from trade unions... And it came from regulators.”
Extractive institutions concentrate wealth while inclusive institutions distribute power and opportunity; technology amplifies whichever institutional form exists.
- Institutions—the formal and informal rules societies build—come in two broad types: extractive institutions, both economic and political, are designed or evolved to enable a small group to extract as much as possible (labor, assets, minerals, financial resources) with little constraint, as under the Stuart monarchs; inclusive institutions instead empower many people through secure property rights and opportunities to pursue any occupation, and inclusive economic institutions are hard to sustain without inclusive political institutions that distribute political power broadly.
“they are institutions that are designed or have evolved in such a way that enable a small group to extract as much as they can.”
- Occupational restriction is a form of extraction: under South African Apartheid, the color bar explicitly prohibited Black South Africans from being anything but unskilled laborers in mines, agriculture, or factories—barring them from becoming bricklayers, plumbers, or supervisors—thereby forcing people into the lowest-paid, lowest-skill occupations, whereas an inclusive economy enables people to pursue their skills.
“The color bar explicitly prohibited black South Africans from becoming, for example, bricklayers or plumbers or supervisors. So that is another form of extraction”
- Although wealth inequality is hard to estimate because some forms of wealth are not on tax forms, economists who study it believe wealth inequality is now much higher than at any point in at least the 20th century—exemplified by Elon Musk's $56 billion pay package—confirming that modern technology has been making extraction more efficient and concentrating wealth.
“Economists who work on estimating wealth inequality think that it's much, much higher than it's ever been, at least in the 20th century.”
- When Adam Smith wrote The Wealth of Nations, the gap between the richest and poorest nations was roughly fourfold, whereas today it is about sixtyfold, showing the world has become dramatically more unequal across nations—a divergence that Why Nations Fail argues stems from extractive versus inclusive institutions and is entangled with politics, economics, and power.
“When he was writing, probably the gap between the richest and the poorest nation was about fourfold or something. Today that's like 60 fold.”
Worker voice and knowledge are essential for designing genuinely productive technology and distributing gains equitably.
- A large fraction of the economy involves social interaction, real-world action, wisdom, judgment, and creativity that are beyond current AI; but if AI is treated as an informational tool that provides context-dependent, reliable, real-time information rather than something that mimics humans, it can raise the marginal productivity of many workers—for example, helping an average electrician troubleshoot a complex problem as well as the very best electrician could.
“if you think of AI as an informational tool, not as something that can write Shakespearean sonnets or pretend to be human. But as an informational tool that provides to you context dependent, reliable real time information, it can actually make many people more productive.”
- Worker voice is needed not only for an equitable division of gains but also to know what technology to build, because much of the implicit, tacit knowledge of what works, what doesn't, and where better information is needed resides with the workers themselves; tech leaders designing tools in isolation without ever talking to workers would not understand the work well enough to make it genuinely more productive.
“A lot of that implicit tacit knowledge is with the workers. They know what works, what doesn't work... So actually you need worker voice, both for a more equitable division of the gains, but you also need worker voice to actually know what to do with our technological capabilities.”
- Correcting the harms of technological change is possible but never self-acting or automatic; hope is not the same as blind optimism, so rather than passively letting the market work and assuming things improve, society needs a proactive agenda—better institutions to protect the disadvantaged and deliberate steering of technology to make workers more productive and to create better information environments rather than filter bubbles.
“hope is not the same as blind optimism... it wasn't self acting, that wasn't automatic.”
Business models exploiting social psychology and data create harms; alternatives and regulation can reshape digital incentives toward public benefit.
- While Henry Ford recognized the need for a prosperous middle class to buy cars, extreme inequality does not bring the economy to a standstill because resources are redirected toward luxury goods and services demanded by the very wealthy—such as expensive personal trainers and the fast-growing occupation of sommelier—so the market system itself will not self-correct inequality through a lack-of-demand channel.
“you can redirect resources towards producing the goods and services, especially services that the very wealthy will demand. So I don't think just the market system itself will create the forces to limit inequality through that channel.”
- The harms of social media—emotional outrage, extremism, and mental health problems from monetizing the data of hundreds of millions of people—stem from a business model built on selling digital advertising, which requires keeping people glued to their screens by exploiting their social psychology; alternatives like Wikipedia and Netflix do not rely on emotional exploitation, so a digital ad tax could open space for better business models, alongside antitrust, to create a level playing field as was done against Standard Oil in the Progressive Era.
“We've come up with a business model where we sell you digital advertising, and that requires that people are glued to their screen. And the best way of doing that is to play with people's social psychology.”
- Against the materialist view that ideas are mere window dressing on underlying material conditions, ideas genuinely matter and can drive historical change: 17th-century Britain went from the divine right of kings and the great chain of being to, by the 1688 Glorious Revolution, a critical mass favoring constitutional monarchy and constrained government, driven by competing ideas (diggers, levelers, Hobbes, Locke) rather than by material change alone.
“I am on the side of ideas, and this is one of the things that is very difficult to formalize”
Elites justify extraction as serving the common good, so skepticism toward tech leaders' claims of disrupting for social benefit is warranted.
- The British Industrial Revolution was driven not by the highest aristocrats but by the 'middling sort'—often children of artisans or teachers—who, amid scientific discoveries and shifting cultural norms, felt entitled and deserving as the smartest and most talented; while revolutionary in tearing down existing hierarchy, they simultaneously looked down on workers who were not making the discoveries, a sociology that echoes today's tech elite who feel they are the smartest.
“It was what the British called the middling sort of people, not the cream of the crop, the upper crust, not the meaner sort, the middling sort. So many of them were children of artisans or teachers.”