More than 2,000 years ago, Aristotle formulated one of history’s most remarkable thought experiments. What, he asked, would happen if tools could obey commands, even anticipate them? What would happen if they could carry out their work entirely on their own?

“If every instrument could accomplish its own work, obeying or anticipating the will of others,” he wrote in “Politics” around 350 B.C.E., “if the shuttle would weave and the plectrum touch the lyre without a hand to guide them, chief workmen would not need servants, nor masters slaves.”

For centuries, this passage was understood primarily as a philosophical allegory. Today, what is striking is not Aristotle’s technical imagination but his political imagination. He was more interested in the social consequences of self-operating machines than in their feasibility.

Artificial intelligence forces us to confront this issue anew: What becomes of a society whose economy, purpose and welfare state are built on wage labor if wage labor loses that foundational role? Perhaps the true AI revolution will begin in our conception of what it means to be human, not inside Silicon Valley’s data centers.

More Or Less?

What’s better: more people or less work for humans? Few contradictions capture our historical moment more precisely. Never before has a society simultaneously spent so much money attempting to compensate for labor shortages while investing so much in technologies designed to eliminate labor altogether.

If current fertility rates continue and net migration falls to zero, United Nations projections suggest that many Western populations will shrink by about 30% by the end of this century. In countries such as Italy, the decline approaches 40%-50%. Governments have responded with generous family subsidies and large-scale immigration programs but have seen limited economic success. Research suggests that immigration’s fiscal contribution depends heavily on education, employment, age at arrival and migration motive. Highly skilled migrants, the minority, tend to generate net economic gains, whereas low-skilled immigration, the majority, can impose a net fiscal drag on advanced welfare states.

In Britain, falling fertility, population aging and migration have become inseparable topics. Parliamentary debates frequently frame birth rates as a long-term growth problem. At the same time, the government seeks to reduce reliance on migrant labor by expanding domestic training programs. In the United States, similar arguments are made from the opposite side: Economists warn that the native-born labor force is shrinking and that immigration now supplies much of America’s labor-force growth. Also, a new Silicon Valley pronatalism has emerged. Elon Musk, Peter Thiel and other technology leaders warn that collapsing birth rates threaten civilization, even as their companies build systems designed to reduce the economy’s dependence on human labor.

In addition to trying to stabilize the size of their labor force, governments invest aggressively in artificial intelligence to stay internationally competitive. The labor displacement that these technologies will ultimately produce is often ignored or dismissed with reassuring promises that new jobs will inevitably emerge, as they always have in the past.

“Perhaps the true AI revolution will begin in our conception of what it means to be human.”

The management of this contradiction is even more striking in East Asia.

South Korea, with a fertility rate of around 0.75 children per woman, is experiencing a demographic collapse unprecedented in peacetime. The government spends tens of billions each year on birth and family incentives and housing programs designed to reverse the trend. So far, little success. Simultaneously, South Korea has become the most robot-intensive economy in the world with over 1,000 industrial robots per 10,000 manufacturing workers.

Japan presents a similar picture. Its working-age population has been declining for more than two decades, and the country has responded by pursuing one of the world’s most ambitious automation and robotics strategies. Today, Japan possesses one of the highest densities of industrial robots with more than 400 per 10,000 manufacturing workers, compared to the world average of around 120-150.

South Korea and Japan offer a preview of the intelligence economy. Their governments are investing in automation to compensate for shrinking workforces while trying to arrest population decline to preserve domestic demand and the fiscal foundations of their welfare states. Both policies make sense within the industrial economy. But what happens when AI and robotics reduce the economy’s dependence on human labor? Could demographic decline and automation eventually become complementary rather than opposing forces?

The Invention Of Homo Faber

The labor society as we know it today is not an anthropological constant. For most of human history, people were not primarily Homo faber — the productive individual who derives social worth from economic work. They became so only with the rise of modernity, and above all through the Industrial Revolution.

In classical antiquity and most of medieval Europe, manual labor was regarded as a necessity performed largely by slaves and the lower classes. The privileged classes devoted themselves to politics, philosophy, religion, war, arts and leisure. As Hannah Arendt observed in “The Human Condition” (1958), modernity radically overturned the classical hierarchy of human activities. Labor, once regarded as a necessity of life rather than its purpose, increasingly became the organizing principle of society.

Adam Smith provided the intellectual economic foundation for this transformation. In “The Wealth of Nations” (1776), he argued that the true wealth of a nation lies in the total productive labor of its people, not in gold, land or trade surpluses.

Around 1800, roughly 75% of Americans worked in agriculture. Today, fewer than 2% do. America’s industrial transformation created an entirely new economic logic, one symbolized by Henry Ford’s famous wage increase in 1914. Ford understood that his workers should build automobiles and earn enough to buy them. That insight captured the genius of industrial capitalism: production, wages and demand were bound together.

“If artificial intelligence assumes an increasing share of productive labor, society will have to find answers to how to cultivate lives that remain meaningful outside employment.”

This model was at the core of Western democracies throughout the 20th century. Wages financed taxation, welfare systems and pensions. At the same time, they generated income that sustained consumption and served as the principal mechanism of social integration. In numbers, labor accounts for up to 60% of GDP and private consumption for nearly 70% for OECD nations.

The importance of wages extends far beyond the hours people spend at work. Every productive hour finances about nine hours of non-working life: the years of childhood, education and retirement; evenings, weekends and illnesses. Another way to look at it is that every hour of paid work supports those hours of non-working life across society, including workers’ own time outside employment and that of children, students, retirees, homemakers and other dependents. Historically, these ratios have increased, driven mainly by reductions in working hours and increases in life expectancies.

When jobs become scarcer or wages lose their economic bargaining power, this architecture begins to fracture. This puts immense pressure on public finances: Children, retirees, students, the sick and much of public spending ultimately depend on the wages of the working population.

The defining economic problem of our century so far has been How does a society finance ever more years of life with ever fewer years of work? For decades, advanced economies have managed to answer that question through a fragile combination of economic growth and rising public debt. But what happens when artificial intelligence begins to undermine the labor market on which that answer rests?

When Intelligence Becomes Cheap

Every major technological revolution transformed a scarce factor of production into an abundant one. The Industrial Revolution dramatically reduced the cost of mechanical energy with fossil fuels. Globalization in the 1990s and 2000s lowered the cost of physical labor by integrating China and other emerging economies into global production networks. The AI revolution may now do the same with intelligence. Cognition, coordination, pattern recognition and many forms of decision-making could become abundant rather than scarce resources.

John Maynard Keynes anticipated this possibility almost a century ago. In his 1930 essay “Economic Possibilities for our Grandchildren,” he warned of what he called technological unemployment: the possibility that humanity might discover ways of saving labor faster than it could discover new uses for it.

For decades, economists largely dismissed this concern in two ways. The first was Joseph Schumpeter’s theory of creative destruction. Technological progress destroys existing occupations, but it also creates new industries, new professions and ultimately more productive forms of employment. The second was historical experience. Every previous technological revolution had ultimately generated more people employed, not mass unemployment.

With AI, this historical adjustment mechanism may no longer be applicable.

The Stanford economist Erik Brynjolfsson has described this possibility as the “Turing Trap.” Previous generations of machines primarily replaced physical labor in a few defined routines. Human beings remained one step above the machines, becoming their planners or problem-solvers. That historical shock absorber is now shrinking. AI is increasingly performing many of the cognitive activities that once enabled workers to move up the value chain.

MIT’s Daron Acemoglu arrived at a similar conclusion from a different direction. He argues that the key issue is how firms choose to deploy AI. If businesses use AI primarily to replace workers rather than to augment their productivity, the historical pattern of technological progress creating more employment than it destroys might no longer hold. The structure of the emerging AI economy already reflects this logic. Many of today’s most valuable AI companies generate extraordinary market value with remarkably small workforces.

Agentic AI further accelerates this process. Agents no longer automate isolated tasks; they increasingly perform entire workflows: analyzing information, correcting errors, adapting strategies and coordinating complex processes with minimal human intervention.

“For the first time since the Industrial Revolution, labor could cease to be the economy’s scarcest factor of production.”

For companies, the decisive issue is simple: Can AI perform economically valuable tasks at substantially lower cost than humans? Increasingly, for many roles, the answer is yes.

The substitution of labor by machines is unfolding in two waves.

The first affects white-collar work. A junior financial analyst, lawyer, management consultant or software engineer in London or New York may cost an employer 80-100 per hour. Modern language models can already perform many of their tasks for a fraction of a cent. Systems possess another structural advantage. While the knowledge accumulated by an employee spreads only gradually through an organization and often disappears when that employee leaves, improvements made by software are retained and deployed instantly and at scale across an entire system.

The second wave is now reaching the physical economy and the blue-collar population. Humanoid robots developed by companies such as Tesla and China’s Unitree are learning increasingly complex movements through observation and simulation. Their manufacturers hope eventually to reduce operating costs to below $2 per hour. This is in stark contrast to 30-50 per hour for an average American worker.

Research by the International Monetary Fund suggests that nearly 60% of jobs in advanced economies will be affected by artificial intelligence, with almost half facing substantial disruption or substitution. Against that background, a reduction of roughly 30% in total employment over the coming decades no longer appears implausible. Whether the final figure is 20% or 40% is ultimately of secondary importance.

What matters is that, for the first time since the Industrial Revolution, labor could cease to be the economy’s scarcest factor of production. If that is true, the consequences extend far beyond the labor market. They force us to reconsider one of the deepest assumptions of the industrial age.

Shrinking Rich

It is precisely at this point that a possibility emerges that remains almost unthinkable in the discourse today: Population decline, discussed almost exclusively as an economic catastrophe, may prove to be an essential corrective during the intelligence revolution. I call this the 30/30 equilibrium.

The idea is simple. Imagine an advanced economy whose population gradually declines by around 30% over several decades while artificial intelligence and robotics reduce labor demand by a similar magnitude. These developments will not occur simultaneously or at the same pace, yet over time they could broadly offset one another. Fewer people require fewer jobs. As a result, the historical ratio between productive labor and the total population could remain remarkably stable without placing an additional burden on public finances.

The 30/30 equilibrium is, first and foremost, a theory of macroeconomic stabilization in which demographic decline ceases to be an economic disaster and becomes part of the adjustment mechanism. However, it does not create prosperity.

Whether a shrinking society also becomes richer depends on a second variable: capital deepening. As the Solow-Swan growth model demonstrates, long-run prosperity depends less on the size of the labor force than on the amount and productivity of capital available to each worker. If AI and robotics increase productive capital faster than the labor force contracts, output per person continues to rise even while the total workforce declines.

If we want an economy’s GDP to stabilize, we need a booster. A 30% decline in the labor force would have to be offset by roughly a 45% increase in capital productivity. This would raise GDP per capita by more than 40%. Given the learning curves already observed in semiconductors, computing power, artificial intelligence and robotics, together with the widening cost advantage of machine intelligence over human labor, such productivity gains no longer appear particularly ambitious.

Now, suppose instead that the population remained constant while productive capital increased by the same 45%. Total output would then expand by roughly 20%. Conventional economics would celebrate this as the superior outcome.

William Stanley Jevons observed in the 19th century that more efficient steam engines did not reduce Britain’s coal consumption. They increased it. As energy became cheaper, new applications emerged, further expanding demand.

Efficiency produced abundance, but abundance also produced greater resource use.  The same dynamic may now apply to intelligence. As AI becomes cheaper, we will almost certainly consume far more of it. This applies to cognitive work as well as to robotics. Silicon Valley celebrates this future as “abundance.” Yet abundance is not necessarily environmentally sustainable. Every additional product made and consumed at a fraction of today’s cost and price still consumes energy and leaves a negative ecological footprint. In addition, AI systems require substantial electricity, computing infrastructure, rare materials and manufactured hardware. Abundance comes at a high cost to the physical world.

“For the first time in history, we may be able to become fewer without becoming poorer. We may even become richer while placing a lighter burden on the natural world.”

From an environmental perspective, allowing the population to decline naturally while increasing prosperity per person may therefore represent a superior equilibrium. For the first time since the Industrial Revolution, the intelligence economy offers us the possibility of choosing such an equilibrium.

This requires abandoning two assumptions inherited from the industrial age.

The first is financial. Modern economies are organized around debt, money creation and asset valuations that all depend on continual expansion. Tomorrow’s economy must become larger than today’s because yesterday’s obligations must be repaid. A “shrinking-rich economy” may ultimately require a monetary and financial system that derives its stability less from ever-expanding debt than from the growing productivity of capital.

The second is geopolitical. “Big is beautiful” belonged to an era in which economic and military strength depended directly on scale. Larger populations meant more workers, more consumers, more inventors, larger armies, bigger markets and greater geopolitical influence. The intelligence economy changes that logic. Human intelligence is scarce, but artificial intelligence expands usable intelligence at an extraordinary pace and cost, making it an abundant resource. Furthermore, technologies such as autonomous drones increasingly erode the traditional military advantages of scale. The same is true for economic influence.

Ecology offers a useful framework. The ecologist Eugene Odum distinguished between pioneer and climax ecosystems. Young forests grow aggressively, consume resources and prioritize speed over sustainability. Mature forests grow little yet contain far greater biomass, complexity, stability and resilience than the pioneer systems from which they emerged.

Human civilization may not be fundamentally different. The fossil-fuel economy was humanity’s pioneer phase. Its success rested on expansion: more people, more energy, more production and more consumption. Artificial intelligence, combined with demographic decline, might allow advanced societies to enter a different stage of development. Technologies such as robotic recycling, AI-managed precision fermentation and closed material loops already point toward an economy capable of creating greater prosperity with fewer material inputs.

The resulting climax society would not be a society of sacrifice. It would be a society that has learned to distinguish growth from prosperity. For the first time in history, we may be able to become fewer without becoming poorer. We may even become richer while placing a lighter burden on the natural world.

The ambition of the 21st century should no longer be to maximize growth but to maximize prosperity: to shrink rich.

Who Inherits The Loom?

If the intelligence economy delivers the productivity gains its advocates expect, it will generate unprecedented wealth. This raises the central political question of the post-labor age: Who receives the surplus?

Under the existing economic order, the answer is clear. The gains will accrue primarily to the owners of capital, not to labor, as ownership determines who has a claim on output. If AI and robotics replace labor as the decisive productive force, then income will shift away from wages and toward those who own the machines, models, data centers, energy systems and platforms through which the new economy operates.

This creates a problem even for the capital owners. Henry Ford understood that workers were producers and consumers. They had to earn wages sufficient to buy the cars they manufactured. The same logic applies to the intelligence economy but in a more unstable form. If machines produce goods while human beings lose the wages with which to buy them, the demand side of the economy weakens. Even the most productive AI economy ultimately depends on customers who can afford to buy what it produces.

Silicon Valley’s preferred answer to this problem is universal basic income (UBI). The idea is simple. The state provides citizens with a minimum income funded through higher taxes, increased public debt or levies on automated production. In the short term, some form of income floor may well be necessary. It could stabilize demand as the population shrinks and cushion the social shock of technological displacement.

But as a permanent settlement, UBI risks becoming a pharmakon: remedy and poison at the same time.

If financed through public debt, UBI injects additional money and purchasing power into the economy without increasing the supply of productive assets, creating inflationary pressure that erodes part of AI’s productivity gains. If financed by taxing production, businesses will ultimately recover much of the additional cost through higher prices, with a similar effect. In both cases, a significant share of the productivity dividend is dissipated rather than translated into higher real living standards.

Economic history offers a warning. In 1795, English magistrates introduced the Speenhamland system, a wage supplement designed to protect rural workers from poverty. It was intended as relief. In practice, it subsidized employers who could continue paying low wages while the public purse made up the difference. It also entrenched the poor’s dependence on local authorities.

The analogy is not exact, but the lesson matters. Income support may ease the transition, but it is unlikely to provide a durable foundation for a post-labor economy.

“What kind of human beings do we wish to become?”

The challenge, therefore, is to distribute the productivity gains of the intelligence economy without eroding them. That requires a financing mechanism that is neither debt-financed nor production-tax-financed.

Here, Karl Marx becomes newly relevant, though perhaps not for the reasons usually assumed. In the “Grundrisse,” he imagined a future in which wealth would no longer depend primarily on immediate labor time, but on the technological appropriation of what he called the “general intellect”: the accumulated knowledge, science and cooperative intelligence of society. Marx’s most contemporary insight was his recognition that knowledge would become the decisive productive force.

Artificial intelligence gives this idea a literal technological form. AI models are not created in a vacuum. They are trained on the accumulated intellectual inheritance of humanity: books, scientific papers, Wikipedia entries, forum discussions, software libraries, images, public archives, open-source code and countless traces of human communication. They also depend on publicly funded research, energy infrastructure, universities, telecommunications networks, roads, water systems and legal protections. The intelligence economy is built by both private firms and a civilizational commons.

Therefore, should the returns from these systems accrue exclusively to private owners, or should society retain a claim on the productive infrastructure that emerged from its collective knowledge?

The 20th century solved the demand problem through wages. Silicon Valley proposes to solve it through transfers. The intelligence economy may ultimately solve it through ownership.

Society’s claim to a universal capital dividend, instead of a UBI, could take the form of ownership, where governments would secure equity stakes, public stakes or sovereign wealth claims in AI infrastructure that depends on public resources, public data and public markets to operate. The proceeds would not flow into general budgets but would instead be distributed directly to citizens as a universal dividend.

Ownership operates differently economically and is not inflationary. It distributes the productivity gains directly at their source, preserving the full efficiency gains of automation while broadening participation in the wealth it creates.

The former wage-earning seamstress would become a shareholder in the loom rather than a dependent of the state. Aristotle imagined a world in which the self-weaving loom abolished the necessity of labor. The intelligence economy gives us the opportunity to decide who inherits it.

The Devil In Paradise

Suppose we succeed. Imagine that artificial intelligence performs most productive labor, demographic decline cushions the adjustment, a new ownership model distributes the gains of automation without sacrificing productivity and advanced societies once again live within ecological limits. Humanity would come remarkably close to Aristotle’s vision of the self-weaving loom. We might finally become Homo curans.

Even then, two deeper challenges remain. The first concerns meaning. More than 70 years ago, Arendt warned of the danger of what she called “worldlessness.” A society organized around work does not automatically become a society of free citizens once work disappears. On the contrary, people might lose the shared public realm in which they experience responsibility, judgment and a sense of purpose. The deeper crisis of the post-labor age may therefore not be unemployment, but meaninglessness.

The anthropologist David Graeber reached a similar conclusion from a different direction. His analysis of “bullshit jobs” suggested that much contemporary employment does not survive because it is socially indispensable, but because modern societies have forgotten how to organize meaning outside paid work.

If artificial intelligence assumes an increasing share of productive labor, society will have to find answers to how to cultivate lives that remain meaningful outside employment.

The second challenge is even more far-reaching. As the world becomes more complex, we increasingly delegate decisions to artificial intelligence. At first, AI merely assists human judgment. Soon it recommends. Eventually, it decides. Every delegation begins as a convenience and may end in dependency. Human judgment gradually atrophies through disuse.

This process is already visible. We delegate navigation to algorithms, communication to recommendation systems, information retrieval to search engines, financial decisions to automated models, partner selection to matching applications and increasingly even legal and administrative decisions to AI-assisted systems. Each individual delegation appears rational. Together, they transform the relationship between human beings and judgment itself.

The logic of intelligent machines is not identical to the logic of human beings. Their purpose is optimization. Ours is judgment. Confusing the two may ultimately prove to be the deepest alignment problem of all.

As these systems increasingly surpass us in specialized domains, the risk is less that they become human than that humans gradually surrender their own capacity for judgment. We may discover too late that we have become second-class decision-makers in a world optimized by machines: Homo algorithmus.

More than 2,000 years ago, Aristotle asked what would happen if the shuttle could weave by itself. Today, that is a reality. But the self-weaving loom leaves us asking: What kind of human beings do we wish to become? Whether the intelligence economy gives rise to Homo curans or Homo algorithmus will not be decided by technology, but by us.