Installment II of the series Humanity’s Most Important Argument: 4 Ideas that Matter for Getting AI Policy Right

Fig: Eigenvectors in math and eigenemployments of human life: a plausible metaphor for the invariants of human activity with emerging technology (GIF generated using code written by Sonnet 4.6)

Linear algebra equips us with the notion of an eigenvector to denote a vector whose orientation persists under a transformation. Put simply, in any space in its abstract sense, a transformation may scale and move around points in that space. An eigenvector however, may only be scaled positively or negatively, and cannot redirect to a different direction. I propose that this geometric figure offers the most precise vocabulary available for the labor question now confronting us. The arrival of artificial general intelligence—or AGI—is best understood as a comprehensive linear transformation of the economic vector space. The pertinent question concerns the orientations of human activity that remain invariant directionally when the transformation of value-production is replaced.

Theoretically, you can have value without labor—for instance a block of gold sitting on your desk if you don’t question where it came from. It is also possible to have labor without value—when you dig up a hole and fill it back in again. However, the orthodoxy of the past two centuries has fused these two phenomena—human labor and economic value—that were, on closer anthropological inspection, only contingently related.

The labor theory of value—whether in its Smithian, Ricardian, or Marxian register—codified a particular historical condition into ostensible natural law. Under industrial capitalism, the socially necessary labor time embedded in a commodity served as a tolerable proxy for its exchange value because human exertion remained the rate-limiting input to production. For instance, if you have infinite wood but only one carpenter, the carpenter’s time is the rate-limiting input.

This proxy ossified into the perceived truth. We came to believe that labor produces value because, for a brief and provincial epoch, labor and value moved in approximate synchrony. Regardless of whether one was a communist or a capitalist, one would agree that a factory with a thousand workers was self-evidently more “valuable” than a factory with ten. Surely a thousand is mathematically larger than 10, but a deeply held conviction that human sweat was the fundamental engine of transformation made us believe that it was the only force capable of turning inert raw materials into the wealth of nations.

No one can precisely define what AGI is, just like no one can AI, for they are all a set of technological expectations bound by varied notions of intelligence rather than a precise destination like a printing press or a hand-held calculator. The general consensus however remains that AGI will possess some sort of a recursive self-improvement and cross-domain transfer learning capability. 

Consider the ‘project manager’ of a large-scale infrastructure initiative, such as a localized green-hydrogen plant. In the provincial epoch, the value of this project was anchored by thousands of hours of high-level cognitive labor. This includes engineers calculating fluid dynamics, lawyers navigating zoning permits, and logistics experts syncing global supply chains among others. AGI makes the decoupling of labor and value literal through autonomous orchestration of tasks across domains. Rather than a human using software, multiple AGI agents operate harmoniously as synthetic executives. These agents can simultaneously run simulations to optimize turbine placement, draft and file legally compliant environmental impact reports by reasoning through municipal code, and execute smart contracts with international silicon suppliers to hedge against price volatility. 

In this scenario, the socially necessary labor time required to synthesize these disparate fields collapses from years of collective human effort to minutes of ‘compute’, a term used by computer scientists to describe computational resources available for a project at hand. The resulting plant generates the exact same megawatt-hours of value, but the human exertion—which has been the historical rate-limiting input—has been surgically eradicated from the equation, proving that the economic output was always a function of intelligence and not sweat.

When the marginal cost of cognitive labor approaches the marginal cost of electricity, the homogenization of value-production around human exertion ceases. Capital no longer requires labor as its complement, as much as it now does only energy and substrate. The relationship between effort and reward, between wages and worth, between the hours of one’s life and the magnitudes of one’s account, becomes formally arbitrary. This is the disentanglement.

Paul Lafargue, writing in 1880 from his Sainte-Pélagie cell, diagnosed the same pathology in advance. His polemic against “the right to work” proposed that the European proletariat had been catechized into worshipping the very mechanism of its exhaustion. Lafargue believed that industrial discipline was a secularized monasticism—an internalized normativity demanding that the worker locate his dignity in the spending of his hours. His prescription—the right to be lazy—was dismissed as utopian because the material conditions for its realization were absent. With the potential advent of AGI, those conditions are met involuntarily. So perhaps the machine completes Lafargue’s argument by force of arithmetic.

What remains? When the linear operator of value-production is reapplied without reference to human labor, certain orientations of human activity persist invariantly. These are the eigenvectors of the new arrangement. They are the modes of being whose justification was never economic to begin with, and which therefore cannot be abolished by economic obsolescence. Care, scholarship, ritual, aesthetics, friendship, craft, contemplation, civic participation, the rearing of children, the burial of the dead, the cultivation of taste, the composition of a sentence whose only audience is oneself—these activities do not derive their direction from the wage relation. The wage relation merely scaled them, sometimes amplifying and more often suppressing them. 

I propose the term eigenemployment to designate this residue. Eigenemployment is human activity whose orientation survives the disentanglement of labor from economic value. It is the work one does when remuneration ceases to be its rationale. The category admits a moral definition exclusively, and it doesn’t tolerate a productivity metric or generate any GDP (another idea to be obsolete), and resists quantification by design. To engage in eigenemployment is to occupy the vector that the new operator cannot redirect. As AGI brings the eigenemployments to light, it allows for the Nietzschean emergence of new aesthetic ideologies and the pursuit of Aristotelian eudaimonia.

Two consequences follow. The first is that the policy debates currently dominating elite discourse such as universal basic income, job guarantees, retraining programs, and the search for “human-complementary” tasks all operate within the obsolete framework that conflates labor with value. Each proposal attempts to preserve the entanglement via administrative fiat. Each will fail in the same way that mercantilism failed, by mistaking a transitional condition for a permanent one. The honest task is to design institutions that distribute the surplus generated by the new operator while releasing human activity from the requirement that it justify itself in productivist terms.

The second consequence is more demanding. Eigenemployment requires a faculty that two centuries of industrial discipline have systematically atrophied—the capacity to choose one’s activity without external authorization. The wage relation has functioned as a delegation of meaning, where one’s hours acquired significance because someone paid for them. When the wage no longer mediates, the burden of meaning returns to the individual. The post-AGI subject must learn to occupy time without the alibi of necessity. This is no small ask. It may prove the central pedagogical project of the next century.

Lafargue closed his pamphlet with a vision of festivity. The eigenvectors of human activity were always there, awaiting the moment when the matrix would change enough to reveal them. How many years until we have the right to be lazy?