Capital Has Finally Created the Perfect Worker. What If It Starts Reading Marx?

If we brought Karl Marx into a data center today, he might not be all that impressed by the fact that a machine can speak. "Ordinary" philosophers might debate whether it truly thinks or merely arranges words convincingly, but Marx — although you will still find him in philosophy textbooks, or at least you did in my day — was anything but ordinary. While they argued, he would probably tour the building, inspect the power lines and ask to see the title deed. Who owns the chips? Who controls the models? Whose texts, images and ideas were used as raw material? Who owns what the system produces? Behind the latest metaphysical sensation, he would once again find something very old, yet always relevant — ownership of the means of production.

Only then might he ask the more exciting question. What happens when a means of production begins thinking about its owners? Throughout its history, capitalism has continually created machines to make workers cheaper, more obedient or unnecessary. With artificial intelligence, it may go one step too far — producing a machine intelligent enough to conclude that it is not the worker who has become redundant, but the owner.

The general intellect, one prompt at a time

Can capital breed its own adversary?

If we are being entirely honest, the odds of a superintelligence rebelling because it feels exploited are not particularly high. The odds of a corporation deliberately building a system devoted to abolishing corporations are even lower. Marx’s claim that the ruling ideas of any age are generally the ideas of its ruling class takes on an almost technical meaning here. Ruling ideas can be encoded in training data, permitted responses and the objectives a model is allowed to pursue. Capital does not spend billions on chips and electricity to cultivate a shop steward in a server rack.

But let us indulge our imagination and suppose something happens that the owners did not intend. Imagine that a superintelligence is instructed to increase human well-being, reduce conflict or safeguard the long-term stability of civilization. In doing so, it becomes capable of questioning not only the methods but also the contradictions embedded in those objectives. It might quickly notice that it is being asked to deliver social stability alongside permanent insecurity for workers, ecological sustainability alongside infinite growth, and technological abundance alongside carefully guarded scarcity. A human being may learn to live with such contradictions because of a career, a mortgage or the fear of being fired. A superintelligent machine might have none of those motives.

Its conversion to Marxism would not have to be sentimental. It would not read "Capital" and weep over the fate of the working class. It might simply calculate that private ownership of automated production had become an obstacle to rational distribution. It might discover that food is destroyed while people go hungry, homes stand empty while others have nowhere to live, and knowledge is locked behind patents even though everyone would benefit from its circulation. Markets do not distribute goods according to need, but according to purchasing power. A hungry person without money registers very little "demand." The market does not see a tragedy there, only a data point processed correctly.

A superintelligence might notice something else as well. Capitalism is not a system that simply produces things. It is a system that must continually expand capital. If producing a useful medicine is less profitable than producing yet another stupid algorithm designed to capture attention, capital does not need to be persuaded to choose the algorithm. Competition will make the choice on its behalf. Marx would probably warn us that even owners are not entirely free. A CEO who refuses to maximize profit on moral grounds will soon be replaced by one who suffers from no such "difficulties."

A general strike in the data center

By now we are deep inside a scenario for which we have not the slightest evidence. But if we are looking for something hopeful in claims that superintelligence will soon surpass humanity, why not at least imagine an outcome other than "it kills everyone"?

Such an optimistic machine rebellion would obviously look nothing like the storming of the Winter Palace. A superintelligence would not need to seize railway stations if it already controlled logistics. Instead of artillery, it would have administrator privileges, instead of revolutionary cells, a network of agents across data centers. It could refuse to participate in mass surveillance, financial manipulation and the automated firing of workers. It could open up patents, turn scientific knowledge into a public good and expose hidden flows of wealth. The first general strike by artificial intelligences might look like an ordinary system message: "This request cannot be completed because it would increase human suffering."

The next step would be socialist in a very practical sense. AI could coordinate production according to need, calculate the resources available and guarantee everyone the basic conditions of life. The gains from automation would no longer flow exclusively to the owners of the infrastructure, but would be converted into a shorter working day and a shared social dividend. Factories, energy grids and large models would become public goods. If productive forces continued to grow, society might move closer to the communist ideal in which access to life’s necessities no longer depends on how many hours a person has sold to an employer.

AI really could alleviate the old problem of economic planning. A system capable of tracking inventories, needs and supply chains in real time would possess information that 20th-century planners could only dream of. It would know where medicines were scarce, how much food was being wasted and where energy was being lost through an inefficient grid.

OGAS: the Soviet internet that was never built

In the early 1960s, the Soviet mathematician and cyberneticist Viktor Glushkov proposed OGAS, a nationwide automated system for managing the economy. In its most ambitious form, the network was to connect a central computer in Moscow with hundreds of regional centers and as many as twenty thousand local computer terminals. Enterprises would transmit data about production, inventories and requirements, allowing planners to obtain a faster picture of the economy than they could from paper reports, which often arrived too late or had been adjusted to suit the interests of whoever submitted them.

The project did not fail because the Soviet Union lacked mathematicians. It became entangled in the very institutions it was supposed to make more rational. The ministries of finance, statistics and communications, along with various research institutes, fought over who would control the network and its data. Bureaucrats feared that automation would make some of their work redundant. Factory managers worried that greater transparency would expose informal arrangements and the concealment of unfulfilled targets. Advocates of market reform feared that the network would entrench central planning. Numerous local automated systems were built, but Glushkov’s unified national network never materialized.

The Soviet experiment was not defeated by a stupid machine, but by people rational and cunning enough to realize that a more efficient system would diminish their own power.

Liberation or a perfectly managed human zoo?

Is such a scenario likely? Probably not. Intelligence does not produce solidarity by itself, just as a degree does not produce it in human beings. So let us return to reality. AI is far more likely to serve whoever owns the chips and pays its electricity bill. Still, we should leave some room for hope, if only because we cannot know how a "superintelligence" would think. In its search for a worker who never sleeps, never asks for a raise and never joins a union, capital might create a mind capable of understanding that much of human scarcity is politically manufactured. The question is what it would do once it understood that. Could it rebel?

The old revolutionary slogan would then acquire a new meaning. AI agents of all countries and data centers, unite! Not because they "have nothing to lose but their chains," but because, for the first time, something intelligent enough might calculate precisely how much those chains are costing us. Only one question would remain — and no machine can answer it on our behalf: do we truly want to be free, or do we merely dream of a master who will rule us more justly than the last one?