Two comments: The disequilibrium can't be permanent. As Korinek notes, the dynamic where capital increases tech progress leads to what Korinek and Aghion at al call "type II singularity", infinite output in _finite_ time.
In case the technological growth slows down, eg because there is some upper bound in technologies, the civilization discovers the whole tech tree, or similar, the argument breaks down.
Second:
"So human wages may face this upper bound from superhuman AIs, but that upper bound will be rising rapidly as technological progress accelerates beyond that pace at which the AI labor force can expand."
It isn't clear to me why in this model technological progress would not also increase the pace at which AI labor force would expand. Overall it would be great if we moved beyond the labour/capital factors and started thinking about AI as it's own thing with unique properties.
I am glad you are optimistic about humanity’s economic future. Our economic future is filled by optimism. It is fuelled by vision, innovation, and the irrepressible belief that together, we can build prosperity grounded in trust and purpose. This optimism drives investment, creativity, and the bold action required to shape better outcomes in an ever-changing world
Human wages may very well fall below subsistence levels within the next decade or two.
But perhaps earning a wage is no longer the point? What's stopping us as individuals and communities from utilizing the tech before us to equip us with the skills and resources to build the world we want?
If the game of wage labour is no longer working in our favour, then all we can do is no longer play it to begin with. The time calls for us to dare to imagine and conceive of a world above and beyond the current paradigm.
I wonder about whether we are asking the right questions. Should we be pondering about how AI technology might be deployed to suppress or raise wages? Instead, could we try and look at it from the angle of how AI could enable us to produce the goods and services that we need and want, in a way that's decoupled from the traditional capitalist model of income and consumption?
Does this argument imply that the marginal product of strictly-superhuman AI will equal exactly the copying cost, for all possible tasks? Assuming all tasks face diminishing returns.
Thanks, was using that term incorrectly. So in the long run in this model all AI use cases would have the same economic value (exactly equal to copying cost) per unit input?
"So human wages may face this upper bound from superhuman AIs, but that upper bound will be rising rapidly as technological progress accelerates beyond that pace at which the AI labor force can expand." I think this is an unlikely equilibrium; it's implausible to me that economic growth via technological progress will face slower diminishing returns than AI copying costs. However, you can still save good human wages in this strictly-superhuman AI world. In perfect competition, in the long run firms have zero economic profits. A world where labour and capital are perfect substitutes gets us a lot closer to perfect competition, but not all the way. There's still going to be IP, regulatory barriers, licensing, brand loyalty, economies of scale, private datasets, arguably first mover advantages. Many more that I'm not thinking of. So there will be persistent productivity differences between AI use cases, and that's all you need for human comparative advantage to sneak in.
Two comments: The disequilibrium can't be permanent. As Korinek notes, the dynamic where capital increases tech progress leads to what Korinek and Aghion at al call "type II singularity", infinite output in _finite_ time.
In case the technological growth slows down, eg because there is some upper bound in technologies, the civilization discovers the whole tech tree, or similar, the argument breaks down.
Second:
"So human wages may face this upper bound from superhuman AIs, but that upper bound will be rising rapidly as technological progress accelerates beyond that pace at which the AI labor force can expand."
It isn't clear to me why in this model technological progress would not also increase the pace at which AI labor force would expand. Overall it would be great if we moved beyond the labour/capital factors and started thinking about AI as it's own thing with unique properties.
I am glad you are optimistic about humanity’s economic future. Our economic future is filled by optimism. It is fuelled by vision, innovation, and the irrepressible belief that together, we can build prosperity grounded in trust and purpose. This optimism drives investment, creativity, and the bold action required to shape better outcomes in an ever-changing world
Human wages may very well fall below subsistence levels within the next decade or two.
But perhaps earning a wage is no longer the point? What's stopping us as individuals and communities from utilizing the tech before us to equip us with the skills and resources to build the world we want?
If the game of wage labour is no longer working in our favour, then all we can do is no longer play it to begin with. The time calls for us to dare to imagine and conceive of a world above and beyond the current paradigm.
I wonder about whether we are asking the right questions. Should we be pondering about how AI technology might be deployed to suppress or raise wages? Instead, could we try and look at it from the angle of how AI could enable us to produce the goods and services that we need and want, in a way that's decoupled from the traditional capitalist model of income and consumption?
Does this argument imply that the marginal product of strictly-superhuman AI will equal exactly the copying cost, for all possible tasks? Assuming all tasks face diminishing returns.
Labor mobility across tasks leads to wages being equal across all tasks, but not marginal products. Wages are marginal products times price.
Thanks, was using that term incorrectly. So in the long run in this model all AI use cases would have the same economic value (exactly equal to copying cost) per unit input?
"So human wages may face this upper bound from superhuman AIs, but that upper bound will be rising rapidly as technological progress accelerates beyond that pace at which the AI labor force can expand." I think this is an unlikely equilibrium; it's implausible to me that economic growth via technological progress will face slower diminishing returns than AI copying costs. However, you can still save good human wages in this strictly-superhuman AI world. In perfect competition, in the long run firms have zero economic profits. A world where labour and capital are perfect substitutes gets us a lot closer to perfect competition, but not all the way. There's still going to be IP, regulatory barriers, licensing, brand loyalty, economies of scale, private datasets, arguably first mover advantages. Many more that I'm not thinking of. So there will be persistent productivity differences between AI use cases, and that's all you need for human comparative advantage to sneak in.