What this covers

Tyler Cowen and Russ Roberts discuss the labor market consequences of artificial intelligence and automation in a conversation that moves from near-term inequality to long-term social restructuring. Cowen's central claim is that machines will bifurcate the workforce: those who learn to work effectively alongside automation will become far more productive and wealthy, while those who cannot will crowd into low-wage service sectors where machines cannot easily substitute for human labor. The conversation ranges across which human skills will remain scarce and valuable in this world, how education will need to change, which regions and demographics will fare better or worse, and whether the resulting inequality will provoke political upheaval or acceptance.

The discussion dwells on three interconnected mechanisms. First, Human-Machine Complementarity—Cowen illustrates this through freestyle chess, where a modestly skilled human paired with a computer beats either alone, but success requires epistemic modesty and skill at Knowing What Questions To Ask rather than raw chess strength. Second, the rising scarcity of conscientiousness: as the world's best education moves online for free, discipline and self-motivation become the binding constraint, favoring self-directed learners over credentialed peers. Third, the political economy of acceptance: Cowen argues no revolution arrives because Americans resent their near peers far more than distant elites, because an aging electorate resists radicalism, and because inequality has risen while crime has fallen, defying revolutionary expectations. The conversation also touches on geographic sorting, the decline of the mid-century middle class, shifts in what people own and value, and the machinery question—drawing a comparison to the lag between the Industrial Revolution's onset and real wage growth, suggesting today's transition will be painful for decades but potentially utopian much further out.

Sharpest takeaway

Cowen argues that smart machines will split the labor market into those who can effectively complement automation and those who cannot, producing a durable bi-modal income distribution where the middle class largely disappears and conscientiousness, motivation, and epistemic modesty become the scarcest valuable human skills.

  • Workers who complement machines (asking the right questions, motivating others) will prosper while those automated away crowd into low-wage service sectors
  • Conscientiousness and discipline rise in value because free educational resources reward the self-motivated
  • The resulting inequality will not spark revolution because of an aging electorate and because Americans resent local peers, not distant elites

The argument · threads12 threads · 38 claims
0.79

Globalization and automation concentrate rewards on top talent while eliminating middle-class job security and wages.

4 pointscentrality 5/5
  • Globalization equalizes wage rates across countries through an arbitrage applied to the price of labor much like the price of apples, so that while global inequality falls sharply, within virtually every country inequality rises and median wages slightly decline, with much of the effect being zero-sum rather than purely positive-sum.

    “there's a kind of arbitrage applied to the price of labor, as there would be to, say, the price of apples.”

  • Over the coming decades most of the middle class will split into either becoming quite wealthy or stagnating in the lower-middle class range, so that mid-20th-century middle-class America will be seen as a historical anomaly produced by the rise of smart machines and globalization.

    “I think what we thought of as middle class America in, say, the 1960s, will be viewed as a historical anomaly. Lots of people will benefit greatly from this change, but not quite everyone.”

  • Smart machines, computers, and software extend the reach of the most productive workers and create global markets that reward top talent, but simultaneous automation eliminates jobs in an economy that lacks the flexibility of past transitions, so labor force participation falls and income stagnates for other classes.

    “automation is replacing some jobs. But we don't have the same kind of flexible economy that we've had in many times past. And I think on net you'll see labor force participation rates are falling, so fewer people are working. And thus we are getting a lot of stagnation with the other income classes.”

  • Sector-specific evidence shows rising inequality: blockbuster books like Harry Potter and Da Vinci Code earn far more than past best sellers while average author advances are down, and top economists like Krugman and Mankiw earn far more than predecessors while more adjuncts replace tenure-track professors.

    “the Harry Potter books, or Da Vinci Code, they earn much, much more than best sellers of the past. But if you look at average advances for authors, those are in fact way down.”

0.71

Technology enables radical productivity gains in education and knowledge work through gamification, immediate feedback, and AI diagnostics.

5 pointscentrality 4/5
  • Gaming is the sphere of modern life where education has truly succeeded, because complex games teach players in steps that are hard enough to be interesting but easy enough to feel like progress, so the educational breakthrough has already arrived in games and the unsolved problem is merely applying that gamified, immediate-feedback design to teaching other subjects.

    “if you ask in what sphere of modern life has education really succeeded, I think it's gaming. These games are incredibly complex... the game itself teaches people how to play.”

  • Driverless cars basically work now and await only legal and regulatory frameworks, and IBM's Watson applied to medical diagnosis will be online within two to five years with rapidly falling prices, delivering high-quality diagnosis to people regardless of location or income.

    “Driverless cars are coming fairly soon. They basically work now. We need to set up a legal and regulatory environment; that will take a while. But the technology works.”

  • There are essentially no new economic theories left to discover; great papers today come not from re-mining existing data sets but from finding or creating new data, as in development economics, and because machines cannot readily find data sets, this will remain a human province as economics becomes a completely empirical profession and old Chicago/UCLA microeconomics becomes a dying art.

    “In my view there aren't really any new theories left in economics... great papers today are mostly: What data set did you find?... It's people working in areas like development economics finding or creating their own data and then doing something creative with it.”

  • Because the world's best education will soon be available online for free, the binding constraint shifts to who is disciplined and conscientious enough to learn from it, so the shadow value of conscientiousness rises sharply and self-motivated learners like determined Indian Coursera students outperform less disciplined elite students.

    “within 5 years the world's best education will be available online and it will be free. Arguably that's already the case. But the question is: Who is there to learn from this? It's the people who are disciplined and conscientious, which is still distinct from just raw intelligence.”

  • A top-level gamified version of a course could cost around $200 million to build, and companies are reluctant to make that investment until they know the result will be accredited and accepted, but multiple competing learning styles—Khan Academy style, gamified, narrative chat, podcast—will eventually all become reality since people learn differently.

    “are companies right now ready to spend $200 million creating a gamified version of, say, an Introduction to Calculus class? I think until they know it will be accredited and accepted, they are reluctant.”

0.59

Humans working with machines outperform either alone when they are epistemically modest and know how to ask the right questions.

3 pointscentrality 4/5
  • In freestyle chess, a practiced human teamed with a computer beats a computer alone even though computers are stronger than humans, but the best teams do not have the best human players because strong players are tempted to override the computer with their own judgment; the winning humans are epistemically modest and excel at knowing which questions to ask and where the programs disagree.

    “The really good human players are too tempted to override the computer and substitute in their own judgment. The best freestyle teams, they are quite epistemically modest, the human or humans involved. And what they are really good at is asking questions.”

  • Contrary to views that robots will only lower wages by competing with humans (Hanson) or that we will become computer uploads (Kurzweil), a large number of humans can always complement robots and become far more productive working with them, though not every human will be good at this, which is itself a major source of the predicted inequality.

    “I would stress much more that humans can always complement robots. I'm not saying every human will be good at this. That's a big part of the problem. But a large number of humans will work very effectively with robots and become far more productive, and this will be one of the driving forces behind that inequality.”

  • In the future scheme dominated by smart machines, wisdom and modesty will become much greater epistemic virtues, because those who can listen to and defer to machine judgment without losing their sense of self will out-compete smart people who try to do too much themselves.

    “wisdom and modesty will become much greater epistemic virtues in the future scheme. I think that's overall a good thing. We should revere those qualities more. And we will have to, looking forward.”

0.59

Geographic, economic, and educational inequality will rise sharply as wealthy people gentrify scarce coastal cities and conscientiousness becomes increasingly valuable.

3 pointscentrality 4/5
  • Wealthy people will gentrify nice coastal cities like Washington DC and San Francisco, bidding up scarce quality real estate that will not be deregulated, while lower-income people move to cheaper areas in the South, Midwest, Texas, and Oklahoma, producing much greater geographic segregation by income in the United States.

    “wealthy people want to live in really nice cities. That they are capable of gentrifying them... People of lower income are moving out; they are going to the Carolinas or wherever.”

  • Because state universities educating most US students feel budget-squeezed, the model of each school supporting a full line of tenure-track faculty is ending, with top schools funding more superstars while the middle thins out; professors will work alongside online programs and the best teachers will be motivational tutors—like high-paid private tutors in Hong Kong and South Korea earning millions—rather than those who know the most economics.

    “the idea that each of these schools will support a full line of tenure track faculty I think is coming to an end... the people who are the best teachers will not necessarily be those who know the most economics but will be more like tutors.”

  • Because men are on average less conscientious than women, at least in the lower tail of the distribution, women already do better in higher education and many other areas not merely from reduced prejudice but by outright out-competing men, a trend that will be magnified as conscientiousness gains value.

    “if you compare men to women on average which group is less conscientious, I think you have to hand that one to the men. At least the lower tail of the distribution. So I think we already see in higher education and many other areas women doing better.”

0.46

Service sectors and coaching will become major growth areas because machines cannot motivate humans and wealthy people demand scarce attention and talent.

3 pointscentrality 3/5
  • Coaching and marketing will be major growth sectors because computers are not good at motivating us, while human role models, stories, and narratives inspire emotionally; and as a large share of the population becomes wealthy, competition for their still-scarce attention makes marketing the single biggest growth sector of the future.

    “Coaching and marketing, they are both forms of motivation, and one thing I think computers are not good at is motivating us.”

  • As driverless vehicles eliminate well-paying jobs like truck driving, displaced workers crowd into service sectors such as yoga instruction and personal service, and because of abundant competition and indirect competition from overseas, wages in those sectors will stagnate except at the very top.

    “There will be more people trying to do things like be yoga instructors or be personal servants. And as those kinds of jobs grow, I don't see the corresponding wage growth.”

  • Where people actually move today—Texas leads net migration with three of the five fastest-growing US cities—reveals that what people most want is cash in their pocket they control, since Texas offers many jobs and cheap housing despite mediocre weather, crime, and public services; this predicts future migration toward low-cost areas rather than amenity-rich expensive ones.

    “the state attracting the most migrants is Texas. Three of the five fastest-growing cities in the United States are in Texas... they have a fair number of jobs and really cheap housing. And that suggests to me what people really, really like is cash in their pocket which they are in control of spending.”

0.46

The current transition period mirrors the Industrial Revolution's 70-year lag between disruption and broad gains, producing mixed outcomes for 20-30 years before eventual utopia.

3 pointscentrality 3/5
  • The Industrial Revolution beginning around the 1770s-80s did not raise English real wages substantially until the 1840s, so even free-market economists like Ricardo worried about the machinery question; by analogy today's transition will be mixed and difficult for 20-30 years but quite utopian 50-70 years out.

    “In England real wages are not going up at a high pace until the 1840s. So there's there 50-70 years of transition, where even free market economists like David Ricardo are writing about the machinery question and then worried about this.”

  • The near future will mix utopia and dystopia: much more wealth and exciting technologies that help people realize emotions and connection, with more people liberated from unhappy jobs, but also higher risk and economic precariousness for many who remain underemployed in part-time or temporary service-sector jobs they keep stringing along.

    “I also see for a lot of people higher levels of risk, a lot of economic precariousness, a lot of people being underemployed in say part time or temporary service sector jobs that they just keep on stringing along. And that worries me.”

  • Long-run elasticities are always greater than short-run elasticities, so in the longer run there is always more adjustment and adjustment matters more, which is why large changes cued by the financial crisis—what Cowen calls the Great Reset—are still working themselves out.

    “long-run elasticities are always greater than short-run elasticities. So, in the longer run there's always more adjustment and adjustment matters more.”

0.44

Rising inequality stems partly from manufactured perception and mislabeling of distinct income problems, while academics resent wealth for status reasons ordinary people do not share.

5 pointscentrality 2/5
  • Poor people are not politically radical—partly from common-sense wisdom that radicalism won't solve their problems and partly from disengagement, since radicalism requires a conceptual apparatus—while the progressive left in the United States, including Occupy Wall Street, is largely an upper-middle-class phenomenon with little resonance among the poor.

    “the progressive left is largely in the United States an upper middle class phenomenon.”

  • For the most part Americans do not resent the wealthy in general; what they resent are people close to them—colleagues, high school classmates, in-laws—so envy blossoms on Facebook by comparison with peers doing better, not against distant Silicon Valley titans.

    “What people resent are the people who are close to them--their colleagues, people they went to high school with, in-laws. For the most part, Americans do not resent the wealthy.”

  • The radical and more analytically productive approach in the technology-and-inequality debate is to write without being too normative—to keep evaluation at a distance and simply think through what things will be like—rather than injecting emotional outrage at every intermediate point, which has become one of the worst features of the debate.

    “the radical thing to do is to write a book which isn't trying to be too normative; just try to think through what will things be like. And by keeping the evaluation at much more of a distance I think we'll actually get further with the analysis.”

  • Part of the perceived inequality problem is manufactured by repeatedly telling people they have one, so they come to believe it; meanwhile genuine jobs and income problems get mislabeled as inequality problems because they sound similar, though they are conceptually quite distinct.

    “I think part of the reason we think we have an inequality problem is we spend a lot of time telling people we have one. So they start to think it's true.”

  • Academics resent the wealthy because two worldviews compete for status—one holding that the richest deserve more status and one holding that the most academically smart deserve more status—creating a local rivalry that ordinary Americans, who admire distant rich people as entertainment, do not share.

    “it's as if there are two worldviews competing. One worldview is: the richest people deserve more status. And the other worldview is: the most academically smart people deserve more status. And there's a clash there.”

0.40

An older, wealthier electorate will preserve benefits for the elderly while underinvesting in the poor, with no revolution occurring despite rising inequality.

2 pointscentrality 3/5
  • If around 15% of Americans become millionaires, they will vote heavily and wield social and political influence, building a coalition with the elderly that preserves benefits for the old (Medicare is safe) while underinvesting in the poor and young (Medicaid is under fire), which is the most likely political future even if not a good one.

    “they'll build a coalition with a lot of the elderly, also preserving benefits for the elderly, and arguably underinvesting in the poor and the young... Medicare is quite safe, and Medicaid is under a lot of fire.”

  • Despite rising inequality there will be no revolution, partly because the population will be older and old people are less likely to rebel and tend to vote for law-and-order candidates, and notably inequality has risen sharply while crime has fallen, confounding those who expect the oppressed to take to the barricades.

    “I know we'll be much older, and old people are less likely to rebel. But they love to vote for law-and-order candidates.”

0.32

Constant measurement and surveillance will become oppressive as nothing is overlooked, while declining physical ownership and changing cognition reshape how people value possessions and remember.

3 pointscentrality 2/5
  • A world where everything is recorded and measured is oppressive because all the mistakes you make get recorded and graded, whereas we are accustomed to most things being overlooked, so society will need to find ways to back out of this constant scrutiny, which will not be easy.

    “There's something oppressive about it. Because all the mistakes you make get recorded; they get measured. And we're not used to that. We're used to a lot of things being overlooked.”

  • There will be a general decline in the desire to own physical stuff—books, homes, collectibles—visible in lower household formation rates and in falling collectibles prices after eBay, which indicate a net desire to get rid of stuff, with future generations owning far fewer physical books than the current transitional generation.

    “There will be a general decline of stuff, including the desire to own a home of your own. And we see this in the data--lower rates of household formation, lower rates of owning stuff.”

  • Google changes cognition by making it easier to forget facts as long as you know how to find them, so people remember fewer facts but hone their ability to define the right keywords; future generations will remember fewer things but be far better at finding information than those who spanned the pre- and post-Internet worlds.

    “With Google it's easier to forget things, as long as you know how to find them... I remember fewer facts, but I've really honed my ability to define the right keywords.”

0.26

Measured economic indicators increasingly fail to capture actual wellbeing as technology reduces costs and enables value realization outside monetary consumption.

2 pointscentrality 2/5
  • Augmented-reality and recommendation devices that tell you whether someone is lying, what to buy, or what food to discard need not be perfectly accurate; they only need to be a little better than what you would do on your own to be valuable, putting all of recorded intelligence at your fingertips if you are willing to listen.

    “the point is not that this is perfectly accurate. It just has to be a little bit better than you would do on your own.”

  • As people economize more effectively and derive more benefits from the Internet at varying costs, measured wages and wealth will become less accurate indicators of wellbeing, vindicating the old religious idea that money is not the main thing but values and how you use what you have are.

    “wages and measured wealth will become less accurate as real indicators of wellbeing. We'll just see a lot more variations in terms of the costs people face and the benefits they derive from the Internet.”

0.20

State university financing is structurally collapsing as Medicaid crowds out education spending, forcing a shift from tenure-track faculty to superstar and online models.

1 pointcentrality 2/5
  • Recent large cuts to state university systems are mostly permanent rather than recession-driven, with state aid declining in absolute and relative terms (e.g., Virginia's share of state university budgets falling from over 30% toward 10-15% within five years), driven largely by Medicaid crowding out other state spending.

    “In my view it's mostly permanent. Not in every state... in Virginia there have been and will be cutbacks in terms of percentage of the state university's budget which comes from the state. So the old average of over 30% will probably end up within 5 years as being 10-15%.”

0.06

Kurzweil's brain-uploading hypothesis is implausible and represents secular religion rather than credible futurism.

1 pointcentrality 1/5
  • Ray Kurzweil's view that within about 50 years our brains will be uploaded into computers is very hard to believe and amounts to a new kind of secular religion.

    “the Ray Kurzweil view is within 50 or so years we'll all become computer uploads; our brains will be uploaded into computers. I just find that very, very hard to believe. I think it's a new kind of secular religion.”