A Threat Bigger than China | MIT Economist David Autor
What this covers
David Autor, an economist at MIT, discusses how artificial intelligence poses a different and in some ways more diffuse economic threat than previous shocks, and argues for deliberately shaping AI as a tool that amplifies human expertise rather than replacing it. The conversation—recorded with Reid Hoffman—traces the contours of this risk by drawing parallels to the China trade shock of the early 2000s, examining how labor markets adjust to major disruptions, and exploring what education and policy might look like if the goal is to extend valuable decision-making capability to people without elite credentials.
The episode covers substantial ground across economics, work, and education. Early sections establish the China shock as a useful but imperfect comparison: sudden, geographically concentrated, it devastated specialized manufacturing towns even though a million displaced workers would barely register spread across a 150-million-worker labor force. The AI shock, Autor argues, will be different—distributed across industries and occupations, often felt by firms as productivity gain rather than pure loss. The heart of the argument centers on a simple economic principle: wages track the scarcity of valuable expertise, not the social importance of work. A crossing guard prevents collisions with identical stakes as an air traffic controller, yet earns a quarter the pay because certification is scarce for one and common for the other. Automation doesn't eliminate jobs so much as make valuable skills abundant. The conversation then turns to how AI might be designed differently: as a collaboration tool that lets non-experts filter and apply AI output, versus automation that deskills workers by removing the expert portions of their tasks. The final sections address education, inequality, demographic challenges, and why labor markets—despite their flaws—remain more intrinsically distributional than systems based on capital or machines.
David Autor argues that AI's economic impact will differ fundamentally from the China trade shock because it will affect occupations rather than regions, increase firm productivity rather than impose pure competitive losses, and presents an opportunity to democratize access to high-skill work—but only if we deliberately design institutions and policies to enable broad skill development rather than concentrate wealth and opportunity among elites.
- AI will displace specific occupations distributed across regions, not hollow out entire geographic manufacturing centers like the China shock did
- Firms will experience AI as productivity gains rather than existential competitive threats, potentially creating different social dynamics than the manufacturing collapse
- The real challenge is ensuring expertise remains valuable and accessible; automation can either deskill work or amplify human expertise depending on how tools are designed
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
A crossing guard and an air traffic controller do fundamentally the same job—preventing collisions—yet crossing guards earn less than a quarter of what air traffic controllers do, because almost no training or certification is required to be a crossing guard; this illustrates that pay tracks the scarcity of required expertise, not social value.
“think of it like a crossing guard versus an air traffic controller... those are the same job... And yet, crossing guards make less than a quarter of what air traffic controllers do... because there's almost no training or certification required to become a crossing guard.”
Labor markets have a natural adjustment rate of roughly 3% per year (a ~30-year career), so a shock spread over 25 years is manageable through retirement and non-entry, but the same displacement occurring suddenly—e.g., autonomous vehicles replacing 2-3 million drivers on a single Labor Day—would be catastrophic; the slow capital replacement cycle naturally buffers such transitions.
“labor markets have a natural rate of adjustment... 3% of people will retire out of anything... If autonomous vehicles come Labor Day this year replaced all long-distance drivers... That would be a very serious problem cuz there are more than... 3 million... people who just do their living in driving vehicles... If it happened over 25 years... that's kind of a manageable problem... it takes decades to replace all that capital”
If half the people doing leisure, hospitality, and janitorial work were extracted from it, the remaining half would get a big pay increase because firms would have to compete for them—low wages in these jobs stem from oversupply, just as barbers' wages rise over time not because they cut hair faster but because they must be compensated to remain barbers rather than do something else.
“if we just extracted half of the people who were doing... leisure and hospitality and janitorial services from that work the... half who remained would get a big pay increase... why do the wages of barbers rise over time? They're not getting any faster cutting anyone's hair... they have to be compensated to be barbers as opposed to being something else”
The labor market has two virtues a capital-based distribution lacks: it is intrinsically more equal because everyone owns exactly one worker (themselves) in a non-slavery society, with about 60% of US income being labor income that goes first to workers; and in a democracy, if most people work it is easy to argue they are contributors who deserve a vote and a share, whereas if wealth pours from machines it becomes harder to justify everyone's share.
“it's intrinsically a lot more equitable... than the capital market because everybody... owns no more than one worker... They just own themselves... 60% of the income in the United States is labor income that goes first to workers... in a democratic society... if most people are working then it's easy for people to say... they have a vote... Whereas if... all the money... comes out of a fountain... then it's much harder to say that everybody deserves their share”
Workers initially displaced by the China shock largely did not move up or out; 20 years later, many remain in relatively low-paid manufacturing or other low-paid work, while it is their children who never entered manufacturing—showing labor markets adjust across cohorts rather than mid-career.
“the workers who were initially in those locations actually have not moved on. They've not kind of moved up or out uh and many of them are still in relatively low-paid manufacturing or other low-paid work.”
We have lost more clerical jobs over the last 30 years than manufacturing jobs, yet no one talks about a 'clerical shock'—because there was never a 'clerical capital' of the US, since clerical workers existed in every industry rather than concentrated in specialized towns.
“we've lost more clerical jobs uh over the last 30 years probably than we have lost manufacturing jobs. But, no one talks about the clerical shock. Uh why not? Well, one reason is there was never a clerical capital of the United States.”
Demand effects can offset automation: getting cheaper and faster at colonoscopies won't make people line up for more, but cheaper/faster software coding meets vast latent demand (every appliance now has embedded software), so coding work may grow even as the skill set shifts—just as 1990s website-building moved from technical HTML skill to design skill.
“if we... got like really really good and cheap and fast at colonoscopies, people still wouldn't be lining up at their proctologist's office... if we get better, cheaper, faster at coding... There's a lot of demand for software... when people started developing websites... that was all about skill in HTML... Now... it's not really a technical skill, it's design”
Americans work far less than they used to—average annual hours fell from about 3,000 a year 120 years ago to about 1,900 today, with the weekend, vacations, later workforce entry, and earlier retirement meaning we work a much smaller fraction of our healthy lives—so the 'overworked society' narrative is a myth and we have handled increased leisure well.
“at the beginning of the 20th century... people worked on average in the United States about 3,000 hours a year. Uh now we work on average about 1,900 hours of the year... we've invented the weekend... people enter the labor force... 16 18 20 25... and then they retire when they have... 20 years of health remaining... That's kind of a myth.”
The last 40-50 years have been the best in human history by key welfare measures—an unprecedented global middle class, China reducing poverty from about 70% to a couple of percent (over a billion people lifted out), and strong growth and improving health in Central/South America and sub-Saharan Africa—progress that people in the West tend to overlook.
“this has been the best 40 or 50 years that humanity has ever experienced... We've never had a... global middle class until now... China has... reduced its poverty level from 70% to... a couple percentage points... more than a billion people... growth in sub-Saharan Africa has been really strong”
The speed at which occupational change occurs matters enormously. Labor markets have a natural adjustment rate—if careers last 30 years, roughly 3% of people retire yearly without layoffs. If an occupation disappears over 20 years, it's manageable; if it disappears in 7 years, it creates serious unemployment. Autonomous vehicles eliminating 3 million truck drivers overnight would be catastrophic, but over 25 years it's manageable.
“labor markets have a natural rate of adjustment. Uh, if you think a career is 30 30 years, let's say, that means kind of 3% of people will retire out of anything, every year, right?...If autonomous vehicles come Labor Day this year replaced all long-distance drivers, right? That would be a very serious problem cuz there are more than, you know, 2 million, I I believe more than 3 million, you know, people who just do their living in driving vehicles. Um, that So, that would be catastrophic, not because autonomous vehicles wouldn't be a good thing, but because that would be so much job loss all at once. If it happened over 25 years, right? That's kind of a manageable problem, right?”
Workers displaced by the China trade shock have not moved up or out 20 years later; many remain in relatively low-paid manufacturing or other low-paid work, indicating that the transition was not just short-term disruption but sustained disadvantage for that cohort.
“20 years later, those places have kind of rebuilt. Uh they're really quite different. But, the workers who were initially in those locations actually have not moved on. They've not kind of moved up or out uh and many of them are still in relatively low-paid manufacturing or other low-paid work.”
People are paid substantially for expertise and know-how in specific activities, and in rich industrialized countries, there is no longer much value to pure physical labor. Expertise requires two things: it must produce a service that people value, and it must be scarce. When automation devalues a skill set, it happens not because no one needs the skills anymore but because a machine can do it better, cheaper, or faster.
“People are paid to a substantial extent, uh, for expertise, their know-how in specific activities, right? In the, in, you know, rich industrialized countries, there just isn't much of a value to just pure physical labor anymore...For expertise to have market value, uh, it needs to have two things. Uh, one is it needs to be, it needs to produce a service that people value, right?...the second thing is it needs to be scarce...the threat that, uh, automation sometimes poses is it can devalue a skill set very quickly, not because no one needs the skills anymore, but because the machine can do it, you know, better, cheaper, cheaper, faster, right?”
Geoffrey Hinton predicted about a decade ago that AI would eliminate the need for radiologists. However, radiologists now use AI extensively and love it, but there are not fewer radiologists—they just do more of what they did before. They have better tools, and much of their work involves communication with patients and other caregivers that AI does not replace, demonstrating the error of assuming superhuman technology capability means expertise is dead.
“Geoffrey Hinton, you know, famously predicted about a decade ago that we would need no more radiologists. And, uh, because it's so obvious that, you know, AI will just do this better. Now, AI is now used by radiologists, they love AI, but they're not fewer radiologists. Uh, they just do more of what they did. Uh, they they they're no more useful cuz they have better tools, uh, and a lot of their work is not just looking at scans, right? It's all of the communication with the patients and the other caregivers and so on.”
Americans work much less than historically: 100-125 years ago people worked ~3,000 hours/year; now ~1,900 hours/year. People entered the workforce at age 10 and worked until death; now they enter 16-40 and retire with 20 years of health remaining. So we work a much smaller percentage of healthy lives, contradicting the myth that modern society is overworked.
“we do much much much less work than we used to. Right? So at the beginning of the 20th century you know 120 years ago 125 years ago people worked on average in the United States about 3,000 hours a year. Uh now we work on average about 1,900 hours of the year right? We now you know we've invented the weekend uh you know we have vacation and so on. Additionally you know people used to enter the workforce as soon as they were physically able right? You know at 10 years old uh and they would work until they died. Right? And now you know in our society people enter the labor force you know 16 18 20 25 they're PhD students to 40 45 uh and then they retire when they have you know 20 years of health remaining.”
Manufacturing job losses from the China trade shock were extremely regionally concentrated—22% of all US manufacturing employment was lost between 1999 and 2007, with cumulative losses of about a third after the Great Recession. This concentration happened because manufacturing itself was specialized and regionally concentrated, with towns like Hickory, North Carolina (furniture capital) and others specializing in single industries that became non-viable overnight.
“22% of all US manufacturing employment was lost between 1999 and 2007. And then I cumulatively about a third uh once we go into the Great Recession...There was a town that was called, you know, that called itself the sweatshop capital of the world. And another town that called itself uh the furniture capital of the world, you know, Hickory, uh North Carolina.”
Adults make career transitions much more successfully through experiential learning than through classroom instruction. MOOCs (massive open online courses) failed because they replicated classroom structure (videos, lectures) while losing the social component, making them worse than actual classrooms.
“we need to figure out how to use these tools to get better at education, at teaching ourselves and learning new skills. And I and I think, you know, one of the things we know about, especially for adults making transitions, is they learn much more successfully kind of experientially than they do going back to classrooms, right?...remember uh like MOOCs, massively online open courseware, right? Supposed to just everyone was now going to be a, you know, a a botnet herder or, you know, an ethnomusicologist or whatever they wanted to be, and they really weren't very successful. Uh and, you know, people don't talk about them much anymore. Why weren't they that successful? Well, they were like they were like simulated classrooms, right? And it's like everything you hated about a classroom and worse, right?”
For expertise to have market value it must satisfy two conditions: it must produce a service people value (data science, not card tricks) and it must be scarce, because if everyone is expert no one is expert and the skill won't pay; automation threatens wages not by eliminating the need for a skill but by making the machine able to do it cheaper and faster, devaluing the human skill.
“for expertise to have market value, uh, it needs to have two things... it needs to produce a service that people value... The second thing is it needs to be scarce. Uh, because if everyone is expert, no one is expert... it can devalue a skill set very quickly, not because no one needs the skills anymore, but because the machine can do it... cheaper, faster”
Over the last 40 years computerization hollowed out skilled production and office work that followed codified rules, pushing workers who would have done that work into low-paid services (food service, cleaning, security, home health aides) that are socially valuable but require little training and therefore won't pay well—the threat of automation is not running out of work but making valuable skills abundant.
“computerization has led to a lot of hollowing out of both production work and office work... it was feasible to turn them into computer code... a lot of the people who would have been doing office work and manufacturing work 40 years ago now find themselves doing services... it's not expert work... that means it won't pay well.”
Technologies are often complementary to expertise, acting as force multipliers that shorten the distance between intention and result and give us superhuman powers; the key question for any worker is whether technology will make their work more expert or instead displace the valuable expertise they hold.
“our technologies are complementary to our expertise, and they are amplifiers, they're force multipliers for expertise because they... shorten the distance between intention and result... how do you get to be on the right side of this equation”
Good quality of life for families and communities requires investment in schools, safe neighborhoods, and opportunity for the next generation, which cannot be delivered through just low prices but require adequate affluence and investment at family and community level.
“one of the things that really defining for, you know, kind of a good quality of life, in addition to the work you do, it's do your kids have opportunity? Right? Are your neighbors neighborhood safe? Are Are these good schools? And And also equally important, like, do people get a fair shake when they start?... Those things are harder to deliver just through low prices. Right? You know, just, you know, good schools, good opportunity, safety. You know, a lot of that depends on basically having a a level of affluence in your family and in your neighborhood that supports those things.”
Geoffrey Hinton famously predicted about a decade ago that we would need no more radiologists because AI would do it better, but the opposite happened: radiologists now use and love AI, there are not fewer of them, and they do more—because much of their work is communication with patients and other caregivers, not just reading scans, illustrating the hubris of assuming superhuman technology means expertise is dead.
“Geoffrey Hinton... famously predicted about a decade ago that we would need no more radiologists... Now, AI is now used by radiologists, they love AI, but they're not fewer radiologists... a lot of their work is not just looking at scans... it's all of the communication with the patients”
The AI shock will differ from the China shock in three key ways: it will not be regionally concentrated, it will affect occupations and tasks rather than wiping out entire industries, and firms will often experience it as productivity-enhancing rather than as a pure negative competitive shock.
“I think there are three very important differences. Uh one is again that regional concentration... Two, the trade trade shock really impacted the viability of industries in themselves... the third thing is um the China trade shock was experienced by US firms as a pure negative competitive shock... AI will be experienced by many firms as productivity increasing.”
The US is already arguably the wealthiest society in history with no real scarcity, yet has many poor people without healthcare, safe housing, or good schools—not because resources don't exist but because we lack a system willing to distribute to people who don't earn it through the labor market—so more wealth alone won't solve distribution, and the US is becoming less generous even as it grows wealthier.
“we are already an incredibly wealthy society... we don't really have any real scarcity here and yet we have a lot of people who are quite poor... It's that we don't have a system where... we really want to distribute that much to people who don't somehow earn it... the US is... not getting more generous as a society even as it's getting wealthier we seem to be getting less generous”
Quality of life depends not just on low prices and material goods but on economic security and reduced uncertainty—whether rent will rise, whether a cancer diagnosis will cause medical bankruptcy—and these uncertainty problems aren't solved by lower prices; inequality also matters less for the gap itself than for whether it produces 'dynastism' where the next generation starts so far ahead that there is no fair chance.
“those aren't really sort of price issues, those are more uncertainty issues... inequality can lead to dynastism where... the next generation... starts so far ahead of the others that... there's no longer... anything close to fair chance in the next... round... those things are harder to deliver just through low prices”
MOOCs (massive open online courseware) largely failed because they were simulated classrooms—'everything you hated about a classroom and worse'—lacking the social engagement and immersion that make education effective; adults learn transitions much better experientially than by returning to classrooms.
“remember... MOOCs, massively online open courseware... they really weren't very successful... they were like... simulated classrooms... it's like everything you hated about a classroom and worse... At least a classroom is a social environment, whereas watching a video is not... adults making transitions... learn much more successfully kind of experientially than they do going back to classrooms”
Uncertainty matters as much as absolute prices in quality of life. Workers face uncertainty about rent increases, medical bankruptcy from illness, job security, and wage stability. Decline of unions and collective bargaining has made workers more vulnerable to these shocks, even though there are some union costs.
“what is important actually about a lot of this is uncertainty. So, if you didn't have the uncertainty that your rent would go up next year. If you didn't have the uncertainty that someone would get cancer and so you would go into medical debt and you declare medical bankruptcy. Like, those aren't really sort of price issues, those are more uncertainty issues. And so, one of the things that I think a lot of people talk about is declines in unions and collective bargaining. They leave workers more vulnerable to past shocks...they do sometimes provide that certainty.”
The China trade shock caused a rapid, regionally concentrated loss of more than a million labor-intensive manufacturing jobs (furniture, textiles, clothing) between 2001 and 2007, devastating specialized towns even though a million jobs spread evenly across a 150-million-worker labor market would be barely noticeable.
“this in a very short order caused a loss of, you know, more than a million manufacturing jobs... that's not the way manufacturing works. It's very regionally concentrated.”
The US is not running out of jobs but running out of workers, due to low population growth, low birth rates, and heavily restricted immigration, which creates a financing challenge because a large retired population needs workers to support its standard of living.
“we're not running out of jobs. Uh we're much more running out of workers. And, uh, this is true in most industrialized countries... it also means you're going to have a large retired population that's expecting has earned a decent... standard of living in retirement, and you need workers to support that”
Demand matters significantly for job replacement. If colonoscopies became very cheap and fast, people would not line up for more of them, but if software coding becomes cheaper and faster, there is enormous latent demand for software—embedded in appliances, microprocessors everywhere—so more software might be written rather than fewer coding jobs created.
“it also depends on how much what demand looks like, right? So, like if we, you know, got like really really good and cheap and fast at colonoscopies, people still wouldn't be lining up at their proctologist's office to get more of them, right? Uh, but it it is the case that if we, um, you know, if we if we get better, cheaper, faster at coding, right? There's a lot of demand for software, right? Like you can't buy an appliance that doesn't have a microprocessor and it doesn't have embedded software running in it.”
We are not running out of jobs but running out of workers, particularly in most industrialized countries with low population growth, low birth rates, and heavily restricted immigration. This creates challenges not just for finding workers but for financing a large retired population expecting a decent standard of living in retirement.
“we're not running out of jobs. Uh we're much more running out of workers. And, uh, this is true in most industrialized countries, uh, where we have low population growth, uh, low birth rates, and now in the United States heavily, heavily restricted immigration. And that creates real challenges, not just because it's hard to find the workers, but it also means you're going to have a large retired population that's expecting has earned a decent, uh, standard, you know, the right to a decent standard of living in retirement, and you need workers to support that”
Over the last 40-50 years, humanity has experienced the best era in its history, with unprecedented poverty reduction (over a billion people in China alone moved from poverty), the creation of a global middle class for the first time, and prosperity increases in Central America, South America, and sub-Saharan Africa.
“this has been the best 40 or 50 years that humanity has ever experienced, right? The amount of people brought out of poverty. Right? We've never had a world a global middle class until now. And and and partly this is China itself, right? China has, you know, reduced its poverty level from 70% to a, you know, effectively a couple percentage points, and that that's more than a billion people, but it's also created prosperity in Central and South America, in sub-Saharan Africa.”
High inequality creates dynasticism where one generation's wealth advantages compound, preventing genuine equal opportunity in subsequent generations. Even if families earned wealth on merits initially, subsequent generations start so far ahead that fair competition becomes impossible.
“inequality can lead to dynastism where the you know, the next generation is like one kid said the kids starts so far ahead of the others that, you know, sure that uh even if they're, you know, their family got wealthy on the merits, then then there's no longer a kind of of, you know, anything close to fair chance in the next in the next round.”
AI is not reliable for tasks you don't understand—it will misunderstand and lead you astray, and may fabricate plausible falsehoods you can't detect—so a key skill is knowing when you know enough to judge its output; AI works best as a collaboration tool in domains where you have enough expertise to filter and interpret what it produces.
“AI isn't really useful for things that you don't understand. Uh because it's not that reliable... if you're using AI to do something you really don't get you're kind of out over your skis... it might very well make up some Roman currency you've never heard of and you wouldn't know... learning the instinct of when you know enough to know if it's doing something useful”
A consensus in the learning research literature is that AI tutors should not simply tell people what to do, because that isn't how we learn; instead, presenting contrasting options ('take A and this would happen, or B and this would happen') gives learners information to reason about, which is far more instructive than a directive answer.
“a consensus in the literature is what they should not do is to sit around telling us what to do... instead... the AI can say... you could take A and this would happen or B and I predict this would happen... the contrast between those things actually very instructive for learning... you need to give them information that supports that choice and enables them to reason about it”
Contrary to the narrative that America is in decline, US productivity has risen about 30% relative to Europe over the last 20 years, US labor markets work well, and the US has an exceptional culture of innovation—so Americans underestimate both their strengths and their good fortune.
“people think that America's been in decline, but you know, our productivity has risen 30% relative to Europe over the last 20 years... Our labor markets actually work really well... the US is an incredibly innovative country”
22% of all US manufacturing employment was lost between 1999 and 2007, and cumulatively about a third once the Great Recession is included.
“22% of all US manufacturing employment was lost between 1999 and 2007. And then I cumulatively about a third uh once we go into the Great Recession.”
Technology is not inherently automating or deskilling. Rather, technology can either increase expertise by eliminating supporting tasks (allowing focus on valuable expertise) or deskill work by automating the expert parts and leaving only lower-skill last-mile tasks. The outcome depends on how tools are designed and deployed.
“an automation, and it's not just automation, can either increase the expertise of your work by eliminating the supporting tasks and allowing you to focus on what you're really good at, right? So, I don't spend time inverting matrices, I can just work on what this stuff looks like Um, or it can deskill your work by automating the expert parts and just leaving you with a sort of last mile, right? Uh, so both are possible.”
The labor market has advantages as a distribution mechanism compared to capital markets because in a non-slavery society, everyone owns exactly one worker (themselves) and starts at relatively even footing. Since 60% of US income is labor income, work-based distribution creates much more shared resources than capital-based distribution. Work also provides identity, structure, and meaning.
“I actually think the labor market has so much going for it. Uh two things. One is it's intrinsically a lot more equitable or equal than the capital market because everybody in a society that doesn't have slavery and doesn't have labor coercion everybody owns no more than one worker right? They just own themselves. And so we all start off you know at a kind of a at a relatively even starting point. Uh and so and so you know 60% of the income in the United States is labor income...work also has a lot of virtues I think you know gives people identity gives them structure gives them meaning.”
The future is not a forecasting exercise but a design exercise—we are building it. Breaking our way to a positive AI future is not a matter of luck but of making good collective choices, which is extremely hard to do but feasible.
“you know, as uh um my friend uh Josh Cohen, a philosopher, you know, likes to say, you know, the future is is not a forecasting exercise, it's a design exercise, right? We're building it. And so, breaking our way is not just a matter of luck, it's a matter of making good collective choices, and that's extremely hard to do. Uh and so, that is what's feasible, uh but not easy.”
Many of the positive AI outcomes are feasible—the issue is not capability but political will. People know AI could be used for great things or terrible things; uncertainty comes not from AI but from uncertainty about what humanity will choose to do with it.
“many of these things are feasible. If we think we're not going to do them, it's not cuz we couldn't do them, it's cuz we're somehow not delivering on what is feasible. And I think that is, you know, And that's the kind of sad thing. Like, everybody knows AI could be used for all these great things. Everybody also knows it could be used for really terrible things. Uh and your belief about what's going to happen is not really a belief about AI, it's a belief about what humanity will do with this opportunity.”
Automation can either increase the expertise of work by eliminating supporting tasks so you focus on what you're good at, or deskill work by automating the expert parts and leaving only a low-skill 'last mile'; it also often creates new work requiring new expertise, but usually performed by different people.
“an automation... can either increase the expertise of your work by eliminating the supporting tasks and allowing you to focus on what you're really good at... Or it can deskill your work by automating the expert parts and just leaving you with a sort of last mile... it also often creates new work that requires new forms of expertise... it's usually different people who are doing that work”
China's WTO accession in 2001 and permanent normal trade relationship with the United States in 2000 induced a surge in exports to the United States because of falling tariffs and changed investment environment, making China more competitive as a result of having to reform and open.
“they become a member of the World Trade Organization. Uh they also get uh permanent normal trade relationships with the United States in 2000. And that induces a a really a surge in exports to the United States. Uh you know, because of falling tariffs, but it changed investment environment in all kinds of ways. China actually became more competitive as a result of having to reform and open because of the WTO.”
Machines acquire skills much more rapidly than people can. Once a machine does something, many machines can do the same thing very quickly. This is a real concern for occupations like language translation, illustration, and software coding where one machine solution can be instantly deployed to many workers.
“machines can acquire skills much more rapidly than people can, right? Once you have a machine that does something, wow, then you have a then you have a lot of machines that do the same thing. And, uh, and there are we will see this...If you're a language translator, right? That's, you know, you're under threat. Uh, if you're an illustrator, you're under threat. Uh, you know, I think, you know, a lot of people who do just sort of, uh, workman software coding, right? There will be fewer”
When websites emerged in the mid-1990s, the valuable skill was HTML markup language, but as websites became ubiquitous, the valuable skill shifted to design and information presentation rather than technical coding, showing how technology can shift the nature of valuable expertise within an industry.
“So, you know, when people started developing websites, you know, back in the mid-90s, right? That was that was all about skill in HTML, right? Writing markup language, you know, they those websites if you go back and look at them, they they're so incredibly horrible looking, it's hilarious how primitive they are. Now, people there's lots of people who build like websites for a living, but it's not really a technical skill, it's design, right? It's how do you present information? It actually has a different skill set that's involved.”
Society's quality of life improvements from technology go beyond just lower prices; they include expanded access and convenience (moving from phone reservations and bank teller lines to instant online booking and digital banking), reflecting genuine welfare improvements even with unchanged underlying costs.
“we use the web this way all the time, right? You know, the the amount of time we spend, you know, waiting on phone queues or trying to buy airline tickets or or dealing with banks and so on. It's you know, it's actually a lot more convenient than it used to be. I I I remember buying plane tickets on the phone. Um the uh or you know, waiting in line uh to see a bank teller.”
Machines acquire skills far more rapidly than people—once one machine does something, you quickly have many machines doing it—so AI-driven displacement of specific roles (language translators, illustrators, workman-level software coders) can happen abruptly; there is already a notable decline in employment among people who write programs, though not necessarily software engineering broadly.
“machines can acquire skills much more rapidly than people can... If you're a language translator... you're under threat. Uh, if you're an illustrator, you're under threat... we do see a big decline in employment in computer coding right now in software development... in the people who write programs”
US economic insecurity for low-wage workers is a policy choice, not a necessary fact of life: someone working at a McDonald's in Norway, Denmark, or France gets vacation, healthcare, and sick leave, showing that security need not be reserved for the middle and upper class.
“people in the United States have much greater level of economic insecurity than... other people in... less affluent economies working the same jobs... If you work in a McDonald's in... Norway or Denmark or even in France, you're going to have vacation... health care... sick leave... that's a choice that we make, but we tend to think it's inevitable, and it's not.”
Healthcare and education—about 20% of US GDP, much of it public money—are where AI offers the greatest opportunity, and investment should focus not just on rare-disease treatments but on making healthcare more available to everyone for longer, higher-quality lives.
“health care and education, two activities... in the United States, that's 20% GDP, a lot of it's public money... this is where there's such great opportunity, where AI could be a tool that could be so helpful... investing doesn't just mean... more treatments for rare diseases, it means things that make health care more available to everyone”
AI will not have the same regional component as the China trade shock because AI will affect jobs and roles across occupations rather than being concentrated in specific industries and regions. Unlike clerical work, which was lost across all industries without creating a 'clerical capital,' AI's impacts will be distributed.
“we don't expect the impacts of AI to have anywhere near that type of uh local impact, right? You know, for example, we've lost more clerical jobs uh over the last 30 years probably than we have lost manufacturing jobs. But, no one talks about the clerical shock. Uh why not? Well, one reason is there was never a clerical capital of the United States. Right? Uh it wasn't that wasn't the way it worked cuz there were clerical workers in every industry.”
Expertise takes a long time to develop, is expensive, and even experts are fallible and variable in quality, so improving expertise development speed would significantly increase welfare by enabling more people to be effective experts.
“Because, of course, it takes a long time. Expertise, you know, it's great, but, you know, it takes a long time to get expertise. It's slow, it's expensive, and even the best experts are fallible, and most people are not the best experts.”
Work that pays well is high-stakes decision-making work requiring discretion (how to land this plane, care for this patient, remodel this kitchen) where simple rules don't exist—because if simple rules existed the work would already be automated.
“work that pays well is decision-making work... where the stakes are high, it's a one-off choice... And there aren't simple rules. If there were simple rules, they'd already be automated”
Much valuable work in advanced economies is monopolized by elites (professors, lawyers, doctors, financiers) protected by high barriers to entry, which keeps healthcare, education, and legal services expensive; enabling more people to compete in these domains via supporting roles would let people move up rather than be pushed down into low-paid services.
“a lot of valuable work in advanced economies is monopolized by elites... we don't face that much competition... there's huge barriers to entry... health care is expensive, education is expensive, legal services are expensive... if we could enable more people to compete in those domains... it would be great if we could use this technology to enable more people to move up”
If even half of the people in low-wage leisure, hospitality, and janitorial services moved into other work, the remaining workers in those fields would receive significant pay increases because firms would have to compete more for workers. Low wages in these sectors result from having too many workers; scarcity drives wages up.
“if we just extracted half of the people who were doing you know leisure and hospitality and janitorial services from that work the the half who remained would get a big pay increase. Right? Cuz firms would have to compete for them more. The problem is there's too many doing it and that's why wages are so low.”
Barber wages rise over time even though barbers don't get faster at cutting hair. The reason is that as productivity rises generally and people have more opportunities, barbers have to be compensated to be barbers rather than doing something else. This illustrates that growing the overall economy and expanding opportunity is what raises service sector wages, not skill development in those sectors.
“why do the wages of barbers rise over time? They're not getting any faster cutting anyone's hair. Right? And the answer is well they have to be compensated to be barbers as opposed to being something else right? So if there's you know in the long run of productivity rises and there's you know there aren't lots of people you if you will need to convince someone to do you know food service or cleaning whatever and you know they may want to they may not but if you want to in a competitive if there aren't that many people available you'll have to pay them more.”
Crossing guards and air traffic controllers perform fundamentally the same job—preventing collisions—yet crossing guards earn less than a quarter of what air traffic controllers earn because there is almost no training or certification required to become a crossing guard, making expertise the determining factor in wage differentiation.
“think of it like a crossing guard versus an air traffic controller. You know, at some fundamental level, those are the same job, right? The job is to prevent, you know, collisions between, you know, people and vehicles or vehicles and other vehicles. And yet, crossing guards make less than a quarter of what air traffic controllers do. And, uh, again, you know, they're protecting our children's lives when they go to school in the morning. So, they're doing valuable work, uh, but they're not going to be paid a lot, unfortunately, because there's almost no training or certification required to become a crossing guard.”
The concern about AI creating mass wealth while most people remain poor is not primarily about insufficient wealth—the US is already an incredibly wealthy society with no real scarcity—but about distribution systems. Even in a wealthy society without scarcity, we have poor people without healthcare, safe housing, and good schools because we don't have systems that equitably distribute resources to those who don't 'earn' them through labor.
“we are already an incredibly wealthy society right? We're arguably the wealthiest society humanity's ever seen and we don't really have any real scarcity here and yet we have a lot of people who are quite poor and don't have access to health care and don't have a safe housing don't have safe neighborhoods don't have good schools right? That's not because those resources don't exist. It's that we don't have a system where people that we really want to distribute that much to people who don't somehow earn it on their own through labor market.”
Education has barely changed in a millennium—if people from 30-40 years ago walked into a classroom today, they wouldn't be surprised by anything, suggesting massive opportunity exists for AI to improve education and teaching by making learning more engaging and immersive.
“I think everyone who looks at AI says, "Oh my god, there's so much potential here for education." It's amazing how little education has changed uh in the last millennium or so. And even, you know, if we walked into a classroom today from Let's say, you know, we all left school 30 or 40 years ago or 20 or we wouldn't be surprised by a single thing in there, right? Or the way things are done, it's really not different.”
Investment in schools and children is critical for long-term equality and opportunity. People in the United States have much higher economic insecurity than workers in other developed countries (Norway, Denmark, France) doing the same jobs, even though these countries provide vacation, healthcare, and sick leave universally.
“I would like to see more investment in our schools, uh in our children. Those things would make a big difference. And then I do think, you know, people in the United States have much greater level of economic insecurity than than do other people in in less affluent economies working the same jobs. And we tend to think it's a it's a necessary fact of life, and it just isn't. If you work in a McDonald's in, you know, uh in Norway or Denmark or even in France, you're going to have vacation, you're going to have health care, uh you're going to have, you know, sick leave.”
Your belief about whether AI turns out well is not really a belief about AI but about what humanity will do with the opportunity; the future is a design exercise, not a forecasting exercise, so breaking our way depends on making good collective choices—which is feasible but extremely hard.
“your belief about what's going to happen is not really a belief about AI, it's a belief about what humanity will do with this opportunity... the future is... not a forecasting exercise, it's a design exercise... breaking our way is not just a matter of luck, it's a matter of making good collective choices”
The US is more productive and innovative than commonly believed. Productivity has risen 30% relative to Europe over the last 20 years. The US has made enormous contributions to innovation (AI, internet era), has functional labor markets, and should recognize both its strengths and opportunities to learn from other models. Many problems like school shootings and healthcare gaps are policy choices, not inevitable.
“our productivity has risen 30% relative to Europe over the last 20 years. Totally. Much much better off. Our labor markets actually work really well...we're pretty we there's a way to lot done a lot of great things...we're also, you know, the US is an incredibly innovative country, right? You know, AI comes from here. So much of the computer the internet era...we should recognize both our strengths and and our good fortune, but also recognize we can learn a lot from uh other models. And it doesn't Not everything that is bad is inevitable. Uh school shootings are not inevitable. Uh and, you know, people not having adequate health care, not inevitable at this level of income.”
Research shows that AI should not tell users what to do (automation), but instead present contrasts—'if you choose A, this happens; if you choose B, this happens'—because showing different outcomes helps people learn through comparison rather than being told the answer.
“And and one thing the a consensus in the literature is what they should not do is to sit around telling us what to do, right? That's not how we learn. Uh and uh instead of they need to interact with us in a way like say, you know, there's one theory uh that says like we get they should, you know, if you're coming to a choice, you know, there's A and B, right? The the AI can say, "Do A." Uh but it could say, "Well, you could take A and this would happen or B and I predict this would happen." And the contrast between those things actually very instructive for learning.”
In a democratic society where most people work, citizens see themselves as contributors worthy of political voice and shared ownership. If resources come from capital rather than labor, it's harder to justify that everyone deserves a share—political claim becomes based on generosity rather than contribution.
“in a democratic society it if most people are working then it's easy for people to say well these are people all contributors to our society of course they have a vote right? You know they of course they they're worthy we're the co-owners of society. Whereas if we're in a world we say well like all the money you know comes out of a fountain in you know San Francisco you know next to the you know Open AI's headquarters or something then it's much harder to say that everybody deserves their share.”
A better world would be one in which more people without a four-year college degree can do software development, legal work, medical technical work, or kitchen design, because AI could let non-elites enter the high-value decision-making domains currently monopolized by the credentialed.
“if a world in which more people who don't have a four-year college degree can do software development, can do some legal work, can do medical technical work, can do kitchen design... That's a better world in my opinion.”
The good AI scenario is one where AI supports people in doing more valuable decision-making work—both using existing expertise more effectively and acquiring it faster—and the great challenge of the era is to build AIs that help people use expertise better and learn faster, which is hard because over-reliance leads people not to bother learning.
“The good scenario would be one where we were able to use AI to support people to do more valuable decision-making work, both to use their expertise more effectively and to acquire it more efficiently... If you rely too much on technology, you kind of won't bother.”
Contrasting the 'Wall-E' scenario (leisure dystopia where no one works) with the 'Mad Max Fury Road' scenario (warlords controlling scarce resources), Autor argues a very wealthy world can still leave most people with nothing, and the Wall-E outcome is actually the good one because the more likely failure mode is violent inequality.
“I like to compare... the the Wall-E and the Mad Max scenario... it's a future where basically people... watch... holographic TV drinking big gulps... But I view that as the good scenario. Uh because the more likely scenario to me looks much more like Mad Max Fury Road where everybody's competing over... a few remaining resources... controlled by... some warlord”
Automation tools (automatic transmissions, elevators, toll takers) fully encode specialized knowledge in machinery and successfully replace jobs, whereas most tools are collaboration tools that require the user to bring expertise—they let you take knowledge you already have and do it faster, further, or better.
“I like to distinguish between what I... call automation tools versus collaboration tools... automation tools are tools... like... the automatic transmission... the elevator... the toll taker... All of these specialized knowledge that was required is now fully encoded in machinery... Most tools require you to bring some expertise to the table to use them”
A central open research question is whether AI tools built to support experts merely yield better results or also help people acquire judgment faster—judgment being the slowly-developed, experiential part of expertise (in law, medicine, carpentry, actuarial work) that makes professionals valuable beyond book knowledge.
“to ask whether tools that are built to support experts... do they help them acquire judgment faster... if you're a lawyer... a doctor... a carpenter... You develop judgment over time. That's part of the expertise that makes you so valuable... You learn how to make the right decisions at the right time”
Well-paying work is characterized by decision-making under uncertainty with high stakes, non-standard one-off choices (like landing a plane, caring for a patient, remodeling a kitchen) that cannot be reduced to simple rules and therefore resist automation, unlike routine codified work which has already been automated.
“work that pays well is decision-making work, right? Where you actually have to, where the stakes are high, it's a one-off choice, right? How to land this plane, how to care for this patient, how to remodel this kitchen, right? Uh, even, you know, how to season this, you know, meal at a restaurant. And there aren't simple rules. If there were simple rules, they'd already be automated, right? Uh, so it actually requires a lot of discretion.”
AI is not useful for things you don't understand because it is unreliable and will misunderstand and lead you astray. Using AI to do things you don't understand is dangerous—you're 'out over your skis.' AI is a good collaboration tool only when applied to things you have knowledge about, allowing you to adjudicate whether output makes sense and ask the right follow-up questions.
“AI isn't really useful for things that you don't understand. Uh because it's not that reliable. Uh and it will misunderstand for you and lead you astray. So if you're using AI to do something you really don't get you're kind of out over your skis. Uh and that's not a good place to be. So that's why I say it's a good collaboration tool because it's complimentary to you know if it's something you know about then you can adjudicate oh this makes sense this doesn't make sense. You can ask the right question and you can kind of you know filter and interpret that knowledge.”
Even if you're a surgeon and an extreme emergency occurs with no other surgeon available, using AI instructions to perform surgery is a last resort, not a primary strategy, because it violates the principle that you should not use tools outside your core expertise domain.
“I could tell an AI you know tell me how to do a you know a surgical procedure on someone. Uh I shouldn't do that. Right? If I'm a surgeon right? And this is an extreme surgical scenario and there's no other surgeon around I need to do something I can like look at the AI give me instructions and I could probably do it. But so you want to use it in the domains where it can collaborate with you and augment you but it it can't automate away substitute for you know just fundamental lack of knowledge in some area.”
A positive AI scenario would be using AI to support people doing more valuable decision-making work, both using their existing expertise more effectively and acquiring expertise more efficiently. The challenge is extremely hard because it requires tools that support learning without doing the work for people, who will otherwise stop trying.
“the good scenario would be one where we were able to use AI to support people to do more valuable decision-making work, both to use their expertise more effectively and to acquire it more efficiently. And I actually think that's one of the great challenges of our era is to figure out how to create tools, AIs, that support people using their expertise better and learning faster. And I think that's very hard cuz it's quite it's cuz the opposite can occur, right? If you rely too much on technology, you kind of won't bother. Uh, and you'll just say, well, it'll it'll do the job for me.”
Autor contrasts the 'Wall-E' scenario (abundant leisure, everyone comfortable but bored) with the 'Mad Max Fury Road' scenario (everyone competing over scarce resources controlled by warlords). He considers Wall-E the better scenario but views Mad Max as more likely given current distribution trends, because abundant resources without equitable distribution creates dystopia.
“So you've all seen the the movie Wall-E right?...it's a future where basically people you know sit around on you know kind of hovercraft arm chairs...they all weigh 300 lb right? And this is supposed to be some sort of future dystopia. Uh because there's no work to do and everyone's bored. But I view that as the good scenario. Uh because the more likely scenario to me looks much more like Mad Max Fury Road where everybody's competing over you know a few remaining resources that aren't controlled by you know some warlord somewhere.”
Workforce transitions historically occurred on industrial timescales with train-then-deploy-for-lifetime model. AI will require constant ongoing adaptation and continuous learning. The question is how to support learning faster so people can transition within a 10-year timeframe rather than waiting for natural generational turnover.
“previously kind of workforce transitions and so forth have happened in what I would think of as industrial time frames. Which is, you know, part of like when you look at our educational system, it's like, you know, train, then deploy for a lifetime. Um versus, you know, kind of a constant recycling. And I actually think that part of what happens as we get into the the new world of AI is that you'll need to have more constant adaptation. It's part of the reason why my very first book that I wrote, Startup of You, was how we're all going to have to be more entrepreneurial in how we think about our work and our careers.”
Simulated environments like flight simulators for pilots and animatronic training dummies for medical students demonstrate effective skill development through engagement and immersion. Similar simulation could work for plumbing, electrical work, and other trades, but it's expensive—however, with AI and VR, this becomes economically feasible at scale.
“I certainly think that there's a lot of skills that people can learn in simulated environments, right? So, you know, like we we do this, right? If you if you want to fly a plane, you're going to spend a bunch of time in a flight simulator. If you if you're uh, you know, learning to do medical procedures, you're going to start off on animatronic dummies that bleed and scream, right? Why don't we do that for, you know, plumbing and electrical work? Well, it's clear, you know, we do it in a few places cuz it's so damn expensive, we don't do it everywhere else, uh even though we could. But we can now.”
Opportunity at Work organization focuses on increasing returns to skill-based work over credentials, recognizing that if people have the skill they should be able to get the job, making skill assessment an alternative to traditional credentialing as a gating mechanism for work.
“we work a lot with this organization opportunity at work which is trying to increase the Yeah. returns to skill based work as opposed to just degrees. If you have the skill you should be able to get the job.”
Inequality matters in two ways: (1) the important way is ensuring basic quality of life (security, health, housing, safety, schools, opportunity for next generation), and (2) the complicated way is relative status gaps between people, with the first being a real policy problem while the second is more ambiguous.
“there's there's kind of two issues that I think get combined in the inequality discussion. Uh one of which I'm extremely sympathetic to and I think is very important to to to navigate. And one of which is complicated. The complicated one is um is like the well, no, I want there not to be that much of a gap between me and you... the one that's not complicated that I think you're highlighting that is extremely important that we solve is the notion of increasing quality of life”
Collaboration differs from automation in that when two experts disagree (answers A and B) and put their heads together, they may arrive at answer C—something neither originally considered—an emergent outcome that cannot happen when a machine simply tells you what to do; this makes collaborative design essential for tools that help people learn.
“collaboration is, you know, like if two people, two experts, look at the same problem and reach different answers... the answer they come up with may not be the average of A or B, it may be C... which is not something that can happen when a machine tells you what to do”
Quality of life includes not just prices and material goods but also control, meaningfulness, respect, personhood, and place in community, with roots in both developing unique skills and reducing costs/uncertainty, and these dimensions are interconnected.
“quality of life isn't just like, oh look, I can you know, I can afford the cheeseburger. That's great. But it's also kind of like, do I have meaningfulness and and control and respect and personhood and you know, place within community and society? And that kind of breaks down into two components. One component is, you know, kind of what are you learning to do that gives you kind of a unique position in your community”
If everything breaks well, AI should enable secure and fulfilling work, expanded access to education, and better healthcare everywhere. These improvements matter not just materially but in terms of opportunity for the next generation and preventing dynasticism.
“if everything broke our way, if we really did this right, uh you know, we would make it's not that we put ourselves out of work, but we would give people more secure and fulfilling work. We would give them more access to education and access to better health care everywhere. And those things alone would kind of improve welfare in so many dimensions, not just, you know, in terms of material standard of living, not just in comfort, but investing in our kids, uh creating opportunity for the next generation.”
China's growth from the 1980s through the 1990s under Deng Xiaoping involved adopting Western techniques, allowing foreign direct investment, and moving hundreds of millions from unproductive rural agriculture into export processing zones, and WTO accession in 2001 plus permanent normal trade relations made China more competitive by forcing reform and opening.
“China just has this incredible productivity growth... They allow foreign direct investment. They allow hundreds of millions of people to move from unproductive rural agriculture into export processing zones... in 2001, they... become a member of the World Trade Organization... China actually became more competitive as a result of having to reform and open because of the WTO.”
Learning when to turn to AI is itself a skill that improves with practice, analogous to how people developed the instinct to Google things; the speaker keeps a chat window open to bounce ideas and uses AI for quick tasks like turning a photographed table into a bar chart in minutes.
“Learning to use AI well, actually, is a... is an important skill in itself... knowing actually when to turn to AI for things is also like not obvious. It's something you get better at... I always have a chat window open and I use it to sort of... bounce ideas or look things up”
A 'counter-life' question—asking people what they would be doing if they weren't doing what they currently do for work—reveals what people are really passionate about and is a better question than 'What do you do for a living?' which overemphasizes work identity.
“I wish people would ask more often about And I ask people this all the time, like, "What is your counter life? What is the thing you would be doing if you weren't doing what you're doing?" Uh and that often tells you about something else that they're really passionate about. And you know, we in America especially we always ask people, "What you do for a living? What's your work? What's your job?" And that so much summarizes people's identity, but uh kind of too much.”
Growth in sub-Saharan Africa has been strong and livelihoods have improved; while still very poor, the region is much less poor than it was with better health outcomes.
“Uh and in fact, you know, growth in sub-Saharan Africa has been really strong. Uh and livelihoods have improved. Like, you know, it's still very poor, but it's much less poor, and health is better.”
People in the West underappreciate the progress achieved and the good fortune of others because they focus on domestic concerns, overlooking how much global welfare has improved.
“And so, you know, we uh we don't appreciate our good fortune, but we also don't appreciate the good fortune of others. So, I think that is overlooked progress.”
Autor's alternative 'counter life' would have been round-the-world sailing, racing, and ocean crewing, indicating passion for maritime exploration beyond his economics work.
“Uh you know, I think there's an another world I could have I could have been a sailor. Uh I love to sail. I still sail. In fact, I'm in a place where I sail. Uh and I could have imagined, you know, doing round-the-world sailing, doing racing, uh crewing.”
On a 7-week cross-country trip after college, the speaker heard on NPR about a computer learning center opening at a Black Methodist church in San Francisco called Computers and You, volunteered there, and became director of education for three years, which sparked his interest in technology, work, and inequality.
“they talked about this computer learning center opening up at a Black Methodist Church in San Francisco uh called Computers and You... I ended up as the director of education for 3 years and that kind of got me interested in technology and work and inequality.”
The Tiny Desk concert series by NPR (started ~20 years ago by music editor Bob Boilen) provides optimism about the future because it creates intimate spaces where diverse music from known and unknown artists is shared, exposing people to music and art they would never encounter otherwise, giving joy and broadening appreciation.
“the answer that's not quite in any of these categories, but it'd have to be the Tiny Desk series of concerts on NPR. I don't know if you guys watch Tiny Desk, but oh my god...this is started like 20 years ago. Bob Boilen, who was their music editor, started inviting bands to come play at his desk. Uh and they would s- sort of film those. And over time this has become an institution...the greatest thing about Tiny Desk is they're in an intimate space, so it's a small group, and it's often people you haven't heard of. And the range of music that you'll encounter is so incredible. Things you would not see...the Tiny Desk concert...are like, you know, 15-20 minute productions, but I look forward to this like so much.”