I really enjoyed this essay. I suspect that we would disagree on a number of things — but that it would be a productive, enlightening disagreement. The “AI wars” are actually driving me to work more on my non-AI papers, because it’s getting so mean and unfair. (Though, also, 3D printing is really fucking cool…)
The core point, about multiple paths of development for new technologies (especially general purpose ones) is correct and important, but I'd like to add an additional factor. Very slowly, the recognition of nonconvexities arising from interaction effects is seeping into economics. It is now more or less the dominant idea in spatial economics, such as urban and international, but it will not stop there. The barrier is the usefulness of models with a single equilibrium for explanation, prediction and advocacy, but the pure logic of interactive nonconvexity is overwhelming.
Path dependence without intervention becomes historical inertia. That's what multiple attractors means in practice. AI tools will be developed to serve existing interests. If we had a different constellation of interests in charge, the tools would be developed for them instead, and those interests would be reinforced or adapted.
Acemoglu is adjacent to this argument, so of course the we-don't-need-no-stinking-collective-action crowd has it out for him.
«Very slowly, the recognition of nonconvexities arising from interaction effects is seeping into economics.»
Since that violates "internal consistency" with the 3 fables/parables of J. B. Clark that is... COMMUNISM! :-)
«The barrier is the usefulness of models with a single equilibrium for explanation, prediction and advocacy»
Those "models with a single equilibrium" are purely for advocacy. So-called reality has optimization landscapes with many local maximums, but Economics does not address "external consistency" but only "internal consistency".
«but the pure logic of interactive nonconvexity is overwhelming.»
It is overwhelmingly likely way to destroy a career in Economics.
It’s so weird how these people think ‘worse bargaining position for labor vis-a-vis capital’ is synonymous with ‘efficiency’. America wastes on the order of 5% of GDP every single year because the labor movement was insufficiently powerful enough to implement a normal healthcare system during the critical juncture of the 1940s and 1950s. https://www.peoplespolicyproject.org/2026/05/26/current-us-healthcare-situation/
Noah is famously an uncharitable freak who thinks everyone else is stupid, but I don’t understand how anyone else thinks giving a tiny percent of people inordinate power will not lead to severe inefficiencies. (E.g., the illegal and prima facially inefficient DOGE actions performed on the whim of a speed-riddled oligarch.)
«so weird how these people think ‘worse bargaining position for labor vis-a-vis capital’ is synonymous with ‘efficiency’»
If productivity is measured in currency, that is dollars of outputs per dollars of inputs, that is not at all weird. That is how mainstream Economists like it. The alternative is to measure it as physical output per physical input (for example how many work-hours and raw inputs are needed to build a car from raw inputs) but that does not allow considering economic rent as a input "factor of production".
"What do you do with a man like that? The only thing that makes sense is to utilize his talents as best you can, and do your best to prevent him from subverting the national interest along the way."
Is there any concrete legislative or policy proposal to prevent Musk from 'subverting the national interest' that Smith hasn't already come out in fierce opposition to? This is about as vague and empty as Dario Amodei claiming that AI needs to be "accountable to the public" in some totally unspecified way that is nevertheless somehow distinct from any actually-existing proposals to regulate AI.
"Unfortunately, however, technological innovation doesn’t fit readily into the standard framework."
Economic theory is not very good at dealing with (the end results) of technologically provoked upheavals and also not with detailed arbitrage processes while moving from one short lived equilibrium to another short lived equilibrium.
It’s dishonest for these people to frame the debate in terms of being for or against technological progress. Technical progress is a given, arguably. The debate is whether progress should be guided and tempered by broad humanistic principles, versus whether it should be controlled by a tiny minority (or maybe just totally unfettered). And obviously tempered vs unfettered is a spectrum, not a binary.
«The debate is whether progress should be guided and tempered by broad humanistic principles, versus whether it should be controlled by a tiny minority»
Ask ordinary people whether if they created some valuable innovation then it should be expropriated for the sake of being "guided and tempered by broad humanistic principles" or it should be their sole property and under their sole control to make as much money as they possibly can for themselves and screw-everybody-else.
That "technological progress" has lifted the living standards of workers is a favourite claim of right-wing propaganda (like that of a book by Gordon) because then they credit "technological progress" to randian heroes of invention and innovation, the so-called "wealth creators".
However 90+% of the increased living standards of workers is rent extracted from the increased use of coal and then oil which have a tremendous "consumer surplus". Invention and innovation have played a role in increasing diffusion and efficiency of use of coal and then oil, but that has been rather secondary compared to simply adopting coal and then oil based machinery.
Wrigley, "Energy and the english industrial revolution (2013): "Noting that one steam horse power was estimated as providing the power equivalent of 21 manual labourers, he suggested that in France ca 1840 steam engines were performing the work of 1.2 million labourers but that by the mid-1880s the rapid expansion in the use of steam engines meant that this figure had risen to 98 million labourers, ‘two-and-a-half slaves for each inhabitant of France’ [19, III, p. 74]."
Andres Malm "Fossil capital" (2016) https://www.versobooks.com/en-gb/products/135-fossil-capital: "One method used by Wrigley and his followers to illustrate the logic is to convert coal into acres of land required to generate the same amount of energy. In 1750, all coal produced in England would have equalled 4.3 million acres of woodland, or 13 percent of the national territory. In 1800, substituting wood for all coal would have demanded 11.2 million acres, or 35 percent of the British land surface; by 1850, the figures had risen to 48.1 million acres and 150 percent, respectively. As early as in 1750, then, a hypothetical total conversion from coal to wood in the British economy would have ‘represented a significant proportion of the land surface for which there were many other competing uses’; in 1800, it would have been ‘quite impractical’; in 1850 – the threshold of 100 percent crossed – ‘self-evidently an impossibility’. [...] Paolo Malanima, likewise standing on the shoulders of Wrigley, estimates that in the absence of fossil fuels, Europe would have needed a land area more than 2.7 times its entire continental surface in 1900, rising to more than 20 times in 2000."
I really enjoyed this essay. I suspect that we would disagree on a number of things — but that it would be a productive, enlightening disagreement. The “AI wars” are actually driving me to work more on my non-AI papers, because it’s getting so mean and unfair. (Though, also, 3D printing is really fucking cool…)
The core point, about multiple paths of development for new technologies (especially general purpose ones) is correct and important, but I'd like to add an additional factor. Very slowly, the recognition of nonconvexities arising from interaction effects is seeping into economics. It is now more or less the dominant idea in spatial economics, such as urban and international, but it will not stop there. The barrier is the usefulness of models with a single equilibrium for explanation, prediction and advocacy, but the pure logic of interactive nonconvexity is overwhelming.
Path dependence without intervention becomes historical inertia. That's what multiple attractors means in practice. AI tools will be developed to serve existing interests. If we had a different constellation of interests in charge, the tools would be developed for them instead, and those interests would be reinforced or adapted.
Acemoglu is adjacent to this argument, so of course the we-don't-need-no-stinking-collective-action crowd has it out for him.
«Very slowly, the recognition of nonconvexities arising from interaction effects is seeping into economics.»
Since that violates "internal consistency" with the 3 fables/parables of J. B. Clark that is... COMMUNISM! :-)
«The barrier is the usefulness of models with a single equilibrium for explanation, prediction and advocacy»
Those "models with a single equilibrium" are purely for advocacy. So-called reality has optimization landscapes with many local maximums, but Economics does not address "external consistency" but only "internal consistency".
«but the pure logic of interactive nonconvexity is overwhelming.»
It is overwhelmingly likely way to destroy a career in Economics.
It’s so weird how these people think ‘worse bargaining position for labor vis-a-vis capital’ is synonymous with ‘efficiency’. America wastes on the order of 5% of GDP every single year because the labor movement was insufficiently powerful enough to implement a normal healthcare system during the critical juncture of the 1940s and 1950s. https://www.peoplespolicyproject.org/2026/05/26/current-us-healthcare-situation/
Noah is famously an uncharitable freak who thinks everyone else is stupid, but I don’t understand how anyone else thinks giving a tiny percent of people inordinate power will not lead to severe inefficiencies. (E.g., the illegal and prima facially inefficient DOGE actions performed on the whim of a speed-riddled oligarch.)
«so weird how these people think ‘worse bargaining position for labor vis-a-vis capital’ is synonymous with ‘efficiency’»
If productivity is measured in currency, that is dollars of outputs per dollars of inputs, that is not at all weird. That is how mainstream Economists like it. The alternative is to measure it as physical output per physical input (for example how many work-hours and raw inputs are needed to build a car from raw inputs) but that does not allow considering economic rent as a input "factor of production".
"What do you do with a man like that? The only thing that makes sense is to utilize his talents as best you can, and do your best to prevent him from subverting the national interest along the way."
Is there any concrete legislative or policy proposal to prevent Musk from 'subverting the national interest' that Smith hasn't already come out in fierce opposition to? This is about as vague and empty as Dario Amodei claiming that AI needs to be "accountable to the public" in some totally unspecified way that is nevertheless somehow distinct from any actually-existing proposals to regulate AI.
"Unfortunately, however, technological innovation doesn’t fit readily into the standard framework."
Economic theory is not very good at dealing with (the end results) of technologically provoked upheavals and also not with detailed arbitrage processes while moving from one short lived equilibrium to another short lived equilibrium.
The Economist does not have anonymous authors. The Economist is written by The Economist.
It’s dishonest for these people to frame the debate in terms of being for or against technological progress. Technical progress is a given, arguably. The debate is whether progress should be guided and tempered by broad humanistic principles, versus whether it should be controlled by a tiny minority (or maybe just totally unfettered). And obviously tempered vs unfettered is a spectrum, not a binary.
«The debate is whether progress should be guided and tempered by broad humanistic principles, versus whether it should be controlled by a tiny minority»
Ask ordinary people whether if they created some valuable innovation then it should be expropriated for the sake of being "guided and tempered by broad humanistic principles" or it should be their sole property and under their sole control to make as much money as they possibly can for themselves and screw-everybody-else.
That "technological progress" has lifted the living standards of workers is a favourite claim of right-wing propaganda (like that of a book by Gordon) because then they credit "technological progress" to randian heroes of invention and innovation, the so-called "wealth creators".
However 90+% of the increased living standards of workers is rent extracted from the increased use of coal and then oil which have a tremendous "consumer surplus". Invention and innovation have played a role in increasing diffusion and efficiency of use of coal and then oil, but that has been rather secondary compared to simply adopting coal and then oil based machinery.
Wrigley, "Energy and the english industrial revolution (2013): "Noting that one steam horse power was estimated as providing the power equivalent of 21 manual labourers, he suggested that in France ca 1840 steam engines were performing the work of 1.2 million labourers but that by the mid-1880s the rapid expansion in the use of steam engines meant that this figure had risen to 98 million labourers, ‘two-and-a-half slaves for each inhabitant of France’ [19, III, p. 74]."
Andres Malm "Fossil capital" (2016) https://www.versobooks.com/en-gb/products/135-fossil-capital: "One method used by Wrigley and his followers to illustrate the logic is to convert coal into acres of land required to generate the same amount of energy. In 1750, all coal produced in England would have equalled 4.3 million acres of woodland, or 13 percent of the national territory. In 1800, substituting wood for all coal would have demanded 11.2 million acres, or 35 percent of the British land surface; by 1850, the figures had risen to 48.1 million acres and 150 percent, respectively. As early as in 1750, then, a hypothetical total conversion from coal to wood in the British economy would have ‘represented a significant proportion of the land surface for which there were many other competing uses’; in 1800, it would have been ‘quite impractical’; in 1850 – the threshold of 100 percent crossed – ‘self-evidently an impossibility’. [...] Paolo Malanima, likewise standing on the shoulders of Wrigley, estimates that in the absence of fossil fuels, Europe would have needed a land area more than 2.7 times its entire continental surface in 1900, rising to more than 20 times in 2000."