In early 2024, Anish Acharya, a general partner at Andreessen Horowitz, a big venture-capital firm based in Menlo Park, posted an article online titled “How AI Will Usher in an Era of Abundance.” Since then, and even before, various Silicon Valley types have been tossing the term around loosely. Last summer, Elon Musk even adopted the term “sustainable abundance” for a new Tesla mission statement. (Over Christmas, Musk substituted “amazing” for “sustainable,” saying the former term was “more joyful.”)
To be sure, abundance isn’t a new concept: it features prominently in the Bible. But it might surprise some people in Silicon Valley to discover that one of the first people to use it in an economic context was Karl Marx. In his “Critique of the Gotha Program,” written in 1875, Marx said that the bourgeois mode of production—capitalism—could be fully transcended only in “a higher phase of communist society . . . after the productive forces have also increased with the all-round development of the individual, and all the springs of co-operative wealth flow more abundantly.”
Like many of Marx’s works, the passage has been interpreted in various ways. But, taken on its face, it seems to suggest that communism isn’t possible until the economy has reached a very high level of output and productivity, which will create new possibilities for organizing society. Of course, a great leap in productivity and economic growth is precisely what many developers and promoters of artificial intelligence claim will happen as A.I. gets even more powerful, and as A.I.-based technologies become more widely adopted. “Double-digit growth is coming within 12 to 18 months,” Musk wrote on X last month. “If applied intelligence is proxy for economic growth, which it should be, triple-digit is possible in ~5 years.”
Musk, Sam Altman, and other tech moguls have self-interested reasons to talk up A.I.’s economic potential: their firms are pouring huge sums of money into it. Economists are forecasting a much more modest impact. Goldman Sachs has predicted that, in the span of ten years, A.I.’s deployment could raise the level of global G.D.P. (not the growth rate) by about seven per cent. For the United States alone, the Penn Wharton Budget Model puts the boost to G.D.P. at just 1.5 per cent. Last year, I wrote about how some surveys showed that companies had struggled to generate much, if any, financial return from the A.I. investments they had made so far, raising fears of an A.I. bubble.
Still, nobody knows for sure what’s going to happen, and it’s worth at least considering the abundance scenario. In an article last year, Boaz Barak, a computer scientist at Harvard who also works at OpenAI, pointed out, “One way to view AI is as injecting to the economy each year . . . a number N(t) of new ‘workers’ that have a certain quality Q(t).” Using a simple equation that related G.D.P. to inputs of labor and capital, Barak calculated that if A.I. created ten million A.I. “workers,” G.D.P. would increase by four per cent. But if there were one hundred million workers, G.D.P. would rise by roughly fifty per cent. These figures were only illustrative: in the real world, Barak pointed out, A.I.’s impact is constrained by the number of jobs that can be automated. But the arithmetic exercise is thought-provoking, nonetheless.
Even if A.I. doesn’t greatly accelerate economic growth, there’s the issue of how it affects employment and wages. The key issue here is whether A.I. primarily complements or substitutes for human labor. If it enables office workers to carry out their tasks more quickly and effectively, for example, it could raise their wages, preserve many existing jobs, and create well-paid new positions for people who are adept at working with A.I. agents. In a recent article, Séb Krier, a manager for policy development and strategy at Google DeepMind, argued that “future workers will likely function as orchestrators of intelligence,” overseeing what A.I. does. Over the longer term, A.I. could also create new jobs and new professions that we can’t currently envision, which is what other transformative technologies have done.
But the fact remains that if A.I. agents can eventually carry out virtually all cognitive tasks without human intervention—a possibility touted by their promoters—many workers could be displaced, and firms may be reluctant to take on new ones. Given the evolving capacities of models like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude, it’s perhaps unwise to wholly discount the prediction from Dario Amodei, the C.E.O. of Anthropic, that within five years A.I. could eliminate half of all entry-level white-collar jobs. Elsewhere in the economy, who knows what could happen? But if the marriage of A.I. and robotics proceeds, in other sectors, along the lines that it seems to be moving in the automotive industry, where autonomous vehicles are already being deployed in some places, taxi-drivers and truck drivers likely won’t be the only blue-collar workers whose jobs are affected.
“It’s clear that a lot of jobs are going to disappear: it’s not clear that it’s going to create a lot of jobs to replace that,” Geoffrey Hinton, one of the pioneers of the deep-learning models that underpin generative A.I., remarked at a conference last month. “This isn’t A.I.’s problem. This is our political system’s problem. If you get a massive increase in productivity, how does that wealth get shared around?” If A.I. abundance does materialize, that will be a central question.
In a recent Substack article, Philip Trammell, an economist at the Stanford Digital Economy Lab, and Dwarkesh Patel, a tech podcaster, pointed out that in standard economic theory deploying more capital raises workers’ productivity and their wages, but reduces the rewards of further capital investment as diminishing returns set in. This “correction mechanism” keeps the over-all shares of income that accrue to labor and capital pretty constant over time. But if A.I. is easily substitutable for labor throughout the economy, and a potential shortage of workers is no longer a bottleneck to production, the stabilization effect disappears, capital incomes “can rise indefinitely,” and the owners of capital receive an ever-growing share of the economic pie, Trammell and Patel write. How far can this process go? “[O]nce A.I. renders capital a true substitute for labor,” Trammell and Patel write, “approximately everything will eventually belong to those who are wealthiest when the transition occurs, or their heirs.”
Trammel and Patel relate their analysis to Thomas Piketty’s book “Capital in the Twenty-First Century,” from 2014, which argued that, under certain conditions, rising inequality is inevitable under capitalism. To address this problem, Piketty called for a global tax on wealth. Trammell and Patel argue that Piketty’s pessimistic analysis hasn’t applied until now, but “he will probably be right about the future.” They also endorse Piketty’s policy solution, writing, “Assuming the rich do not become unprecedentedly philanthropic, a global and highly progressive tax on capital (or at least capital income) will then indeed be essentially the only way to prevent inequality from growing extreme.” (The tax would have to be global, the authors argue, because if capital doesn’t need much labor to produce things it would be even more mobile than it is now, which would enable it to evade national levies.)
The article by Trammell and Patel has already received some pushback online, largely on the ground that its assumption that capital is perfectly substitutable for labor is unrealistic. Brian Albrecht, the chief economist at the Portland-based International Center for Law & Economics, argues that the process of A.I. machines replacing workers is likely to take a long time, and during that transition “standard economic principles apply.” Krier argued that the mere fact A.I. can do something more cheaply or effectively than human workers doesn’t mean it will inevitably replace them. “People pay a lot to go see concerts and Olympic races even if in principle a model can generate the same song and a robot can run faster,” he wrote.
These articles contain some valid points, but neither of them resolves the basic paradox of the abundance economy: as A.I. makes it more productive and potentially richer, it also makes it easier for firms to substitute capital for labor, which could reduce the share of income going to workers, and, in many cases, cost them their jobs. That would obviously be bad for them, and it could also end up being a problem for businesses. If many people don’t have steady wage incomes, who is going to purchase all the goods and services that A.I. agents and robots produce? Consumer spending is the main driver of the economy. “If demand falls, there will be fewer customers, and even the most efficient AI-driven business cannot succeed if people cannot afford to buy its products,” Alex Imas, an economist at University of Chicago Booth School of Business, wrote in another recent online article.
Imas’s solution to this problem is “to make capital ownership more widely shared.” His preferred method of doing this is to set up a sovereign wealth fund, which would take ownership stakes in companies that benefit from the A.I. revolution and distribute dividends to the citizenry, thus providing income and boosting over-all demand. He argues that this would be better than introducing a wealth tax, which could be used to help finance wage subsidies or a universal basic income, because he believes the tax could stifle innovation and growth.
None of these pieces should be regarded as the final word about the A.I. economy, and none of their authors would suggest they are, I’m pretty sure. But they do show how, even in some places not associated with any radical agenda, questions of abundance and the distribution of income and wealth are steadily moving to the center of debates about the economic future, which is where Marx placed them more than a century and a half ago. In the post-capitalist world of plenty that he envisioned, where the ownership of capital had been socialized and everyone shared its rewards, people would no longer have to restrict themselves to one productive activity, such as being a factory worker or a farmer. Someone could “hunt in the morning, fish in the afternoon, rear cattle in the evening, criticize after dinner,” Marx wrote in “The German Ideology.”
Even in the state socialist economies that the Soviet Union and China created, this vision of a wholesale transformation of production and work was never effected, but a non-communist version of it lived on in John Maynard Keynes’s 1930 essay “Economic Possibilities for Our Grandchildren.” While Keynes had no time for Marx and his theorizing, he, too, believed that technology-driven increases in productivity and income would eventually create an economy in which people could spend less time working—say, three hours a day—and the rest of their time following more agreeable pursuits. In this “age of leisure and of abundance,” the love of money as a possession would be “recognized for what it is, a somewhat disgusting morbidity,” and many “distasteful” social customs and economic practices that were retained because they are “tremendously useful in promoting the accumulation of capital, we shall then be free, at last, to discard.”
Keynes’s optimistic vision didn’t materialize, either. But in reading him or Marx, it’s striking how diminished the ambitions of today’s economists and futurists seem by comparison. Regardless of where the debate about A.I. abundance ends up, it’s raising (or re-raising) some fundamental questions about the ultimate purpose of technological progress, and who benefits from it. That, surely, is a good thing. ♦
