But Matt, what's the evidence that textile manufacturing (let's say) is going to be done by a robot who will be cheaper than a third-world worker? All the talk about AI today is about text and image processing; and yes, I'm sure there are people out there doing research on machine learning for other kinds of tasks but is there any evidence out there that warrants this kind of worry?
There is also a huge amount of work getting done on robotics
And the AI we'll accelerate that work and make new things possible
It's over a medium time span, say, twenty to thirty years.I don't see any reason why the vast majority of jobs won't be automated away between AI and humanoid robots
It seems difficult to imagine the AI trajectory in which it *both* essentially ends the demand for labour from rich countries without also creating a situation where poor countries end up ahead. If making a robot that does the work of a human becomes that trivial, they’ll end up ubiquitous everywhere, even if via people giving them away (for instance, this is the mechanism by which many parts of the world that have not succeeded at export-led development have still ended up with an amount of cheap clothing that would be unbelievable to pre-industrial people).
I think the pharma business model is informative here - the US and other wealthy nations command the highest prices thereby subsidizing the cost of development, but if the marginal cost is low enough (as it would be in this theoretical post-scarcity future), selling the marginal robot in a poor nation is still going to make a profit. We don't need to assume any level of altruism for this to work out. Chinese solar panels and EVs seem to be improving quality of life in poor nations too, and for very similar reasons.
I’m actually more optimistic about developing countries being able to bootstrap domestic industries in this scenario. AI advice and some forms of robotics will be relatively cheap and commoditized in this world. Even if elites hoard most of the resource wealth, local communities could still pool resources to get some sort of AI and robotic assistance. This in turn will boost their productivity and allow them to expand their businesses and afford to buy more robots. Such growth will be exponential even if it’s initially quite slow, because each batch of robots purchased will generate revenue to buy even more robots.
Plus, if even a small fraction of western elites donate internationally, that will still be a very large amount of wealth by present day standards, and I suspect a lot of it will go to seeding local industries in poor countries.
One thing that was absolutely crucial to the "ladder" of industrialization that's undersold here: the UK, the US, Japan, Korea - all of these industrial powerhouses had periods of intense protectionism, allowing for the development of infant industries without external competition. This gets left out of the standard story for obvious ideological reasons, but it's a big part of the reason why it's so much harder for poor countries to enjoy the same impacts of industrialization as their historical antecedents - the modern world banking and financial structure and various treaties that dictate tariff behaviors etc. prevent them from doing the same.
Not true. See here on Korea: https://www.aeaweb.org/articles?id=10.1257/mac.20120197. "We develop a model of neoclassical growth and trade in which lower tariffs lead to increased gross domestic product (GDP) per worker via comparative advantage and specialization, and capital accumulation. We calibrate the model and simulate the tariff reductions that occurred between early 1962 and 1989. The model can explain 17 percent of South Korea's catch-up to the G7 countries in value-added per worker in the manufacturing sector. These gains, as well as most of the welfare gains, are driven by two key transmission channels: multistage production and imported investment goods."
An easy way to see why this is dumb is to ask yourself whether it would be better or worse for the economies of poor US states if we had internal tariffs. After all, Mississippi is generally at a lower level of economic development than Massachusetts, has different factor endowments, etc. So wouldn't imposing tariffs on Massachusetts imports benefit Mississippi by protecting infant industries? Obviously, that's a bad idea. Same principle applies to different countries.
I think a lot of this works on the premise of rapid improvements in advanced robotics that I’m not sure we would expect in the modal medium-term outcome. What does the lot of the developing world look like in the scenario in which AI drives major productivity growth (including in goods production) while reducing low-end white collar employment in the developed world, but fails to quickly create robotic agents? I think it’s still probably positive for the developing world.
I cannot imagine the number of times this headline will be clipped by Yglesias haters. 🤣
That being said, while I absolutely believe that AI is going to eat 90% of knowledge jobs, I’m much more skeptical of its ability to reshape industrial labor. Self-driving cars will slowly take over, but the rate of progress is linear, not geometric as AI is. And driving is a lot easier to automate than sewing clothing.
Furthermore, developed countries will have a lot more money to spend on manually intensive goods like clothing, both in quantity and quality. Luxury brands will take off, and my guess is that hand sewn clothing will become a status symbol.
Maybe sewing is an outlier, but I can also see bespoke goods becoming more popular, with sweatshops using AI to customize our now standard products. You won’t just want a Vitamix, you’ll want one built to your own personal specs.
As others have basically pointed out, sweatshops have the least to worry precisely because they pay so little. Those robots cost money, as does the energy and compute to run them. Once the robots come for sweatshops, if they ever do, literally every other manufacturing industry will have been overtaken already.
Let's break this down to first principles. The problem with a country like Nigeria is that whenever it invests in social programs or infrastructure, it faces inflation, primarily in food and energy prices. However, with advancements in AI, robotics, and solar power, we should be able to increase the supply of these essentials by at least threefold. This means that developing countries could expand their welfare programs and fiscal space if the predicted gains in technological productivity materialize without inflation. Having a bunch of young people without any meaningful work can still be problematic in any case
I think humans will have a competitive advantage in service industries for a long time, or perhaps forever. It seems like countries becoming tourist / retirement destinations might replace manufacturing-oriented labor, as getting an actual human to serve you dinner in the United States might become too expensive for most people in the scenario you're describing.
Why wouldn't we expect a Ricardian future where AI is busy automating large-scale and advanced manufacturing, and humans can find profitable work doing anything that's not scalable enough to be more profitable than another automated truck or textile mill? Plenty of things can be exported but still require serendipity or novelty in a way that is hard to automate: think children's toys, or music. And surely others simply won't clear net zero economic profit for the simple reason that AI will be limited by something.
In a lot of ways, this is a good description of the world we already have in Africa: a farmer in Malawi can hardly produce anything for global markets; extractive companies pay off the Congolese government for mineral rights but avoid hiring any of the locals. Is this phenomenon going to expand to places like Bangladesh and Malaysia, or contract, as Africa is predicted to contain 40% of the world's population at the end of the century?
I noticed the Trump-blessed candidate for Texas Ag commissioner is pushing mandatory e-verify, which would pair with the H2B increase but I doubt it would offset it, and I'm still skeptical e-verify will be widely adopted. More broadly skeptical of robotics advances being adequate, but good article.
But Matt, what's the evidence that textile manufacturing (let's say) is going to be done by a robot who will be cheaper than a third-world worker? All the talk about AI today is about text and image processing; and yes, I'm sure there are people out there doing research on machine learning for other kinds of tasks but is there any evidence out there that warrants this kind of worry?
Also some of these jobs are very difficult to automate. Such as steps in textile manufacturing and phone assembly, for instance.
There is also a huge amount of work getting done on robotics
And the AI we'll accelerate that work and make new things possible
It's over a medium time span, say, twenty to thirty years.I don't see any reason why the vast majority of jobs won't be automated away between AI and humanoid robots
It seems difficult to imagine the AI trajectory in which it *both* essentially ends the demand for labour from rich countries without also creating a situation where poor countries end up ahead. If making a robot that does the work of a human becomes that trivial, they’ll end up ubiquitous everywhere, even if via people giving them away (for instance, this is the mechanism by which many parts of the world that have not succeeded at export-led development have still ended up with an amount of cheap clothing that would be unbelievable to pre-industrial people).
I think the pharma business model is informative here - the US and other wealthy nations command the highest prices thereby subsidizing the cost of development, but if the marginal cost is low enough (as it would be in this theoretical post-scarcity future), selling the marginal robot in a poor nation is still going to make a profit. We don't need to assume any level of altruism for this to work out. Chinese solar panels and EVs seem to be improving quality of life in poor nations too, and for very similar reasons.
Yes, I'm much more optimistic about this working out than Matt is.
I’m actually more optimistic about developing countries being able to bootstrap domestic industries in this scenario. AI advice and some forms of robotics will be relatively cheap and commoditized in this world. Even if elites hoard most of the resource wealth, local communities could still pool resources to get some sort of AI and robotic assistance. This in turn will boost their productivity and allow them to expand their businesses and afford to buy more robots. Such growth will be exponential even if it’s initially quite slow, because each batch of robots purchased will generate revenue to buy even more robots.
Plus, if even a small fraction of western elites donate internationally, that will still be a very large amount of wealth by present day standards, and I suspect a lot of it will go to seeding local industries in poor countries.
One thing that was absolutely crucial to the "ladder" of industrialization that's undersold here: the UK, the US, Japan, Korea - all of these industrial powerhouses had periods of intense protectionism, allowing for the development of infant industries without external competition. This gets left out of the standard story for obvious ideological reasons, but it's a big part of the reason why it's so much harder for poor countries to enjoy the same impacts of industrialization as their historical antecedents - the modern world banking and financial structure and various treaties that dictate tariff behaviors etc. prevent them from doing the same.
Idk North Korea seems to do a pretty good job of protecting its infant industries from international competition.
Not true. See here on Korea: https://www.aeaweb.org/articles?id=10.1257/mac.20120197. "We develop a model of neoclassical growth and trade in which lower tariffs lead to increased gross domestic product (GDP) per worker via comparative advantage and specialization, and capital accumulation. We calibrate the model and simulate the tariff reductions that occurred between early 1962 and 1989. The model can explain 17 percent of South Korea's catch-up to the G7 countries in value-added per worker in the manufacturing sector. These gains, as well as most of the welfare gains, are driven by two key transmission channels: multistage production and imported investment goods."
An easy way to see why this is dumb is to ask yourself whether it would be better or worse for the economies of poor US states if we had internal tariffs. After all, Mississippi is generally at a lower level of economic development than Massachusetts, has different factor endowments, etc. So wouldn't imposing tariffs on Massachusetts imports benefit Mississippi by protecting infant industries? Obviously, that's a bad idea. Same principle applies to different countries.
That doesn’t actually refute his point. Are you sure you posted the right article?
No, it does. Look at the timeline of the Korean growth miracle: https://en.wikipedia.org/wiki/Miracle_on_the_Han_River. It starts in the 1960s after the Korean War.
Incredible to see Yglesias write a Noah Smith article, but way better than it has any right to be.
I think a lot of this works on the premise of rapid improvements in advanced robotics that I’m not sure we would expect in the modal medium-term outcome. What does the lot of the developing world look like in the scenario in which AI drives major productivity growth (including in goods production) while reducing low-end white collar employment in the developed world, but fails to quickly create robotic agents? I think it’s still probably positive for the developing world.
I cannot imagine the number of times this headline will be clipped by Yglesias haters. 🤣
That being said, while I absolutely believe that AI is going to eat 90% of knowledge jobs, I’m much more skeptical of its ability to reshape industrial labor. Self-driving cars will slowly take over, but the rate of progress is linear, not geometric as AI is. And driving is a lot easier to automate than sewing clothing.
Furthermore, developed countries will have a lot more money to spend on manually intensive goods like clothing, both in quantity and quality. Luxury brands will take off, and my guess is that hand sewn clothing will become a status symbol.
Maybe sewing is an outlier, but I can also see bespoke goods becoming more popular, with sweatshops using AI to customize our now standard products. You won’t just want a Vitamix, you’ll want one built to your own personal specs.
As others have basically pointed out, sweatshops have the least to worry precisely because they pay so little. Those robots cost money, as does the energy and compute to run them. Once the robots come for sweatshops, if they ever do, literally every other manufacturing industry will have been overtaken already.
Let's break this down to first principles. The problem with a country like Nigeria is that whenever it invests in social programs or infrastructure, it faces inflation, primarily in food and energy prices. However, with advancements in AI, robotics, and solar power, we should be able to increase the supply of these essentials by at least threefold. This means that developing countries could expand their welfare programs and fiscal space if the predicted gains in technological productivity materialize without inflation. Having a bunch of young people without any meaningful work can still be problematic in any case
I think humans will have a competitive advantage in service industries for a long time, or perhaps forever. It seems like countries becoming tourist / retirement destinations might replace manufacturing-oriented labor, as getting an actual human to serve you dinner in the United States might become too expensive for most people in the scenario you're describing.
Why wouldn't we expect a Ricardian future where AI is busy automating large-scale and advanced manufacturing, and humans can find profitable work doing anything that's not scalable enough to be more profitable than another automated truck or textile mill? Plenty of things can be exported but still require serendipity or novelty in a way that is hard to automate: think children's toys, or music. And surely others simply won't clear net zero economic profit for the simple reason that AI will be limited by something.
In a lot of ways, this is a good description of the world we already have in Africa: a farmer in Malawi can hardly produce anything for global markets; extractive companies pay off the Congolese government for mineral rights but avoid hiring any of the locals. Is this phenomenon going to expand to places like Bangladesh and Malaysia, or contract, as Africa is predicted to contain 40% of the world's population at the end of the century?
I noticed the Trump-blessed candidate for Texas Ag commissioner is pushing mandatory e-verify, which would pair with the H2B increase but I doubt it would offset it, and I'm still skeptical e-verify will be widely adopted. More broadly skeptical of robotics advances being adequate, but good article.