The Machine That Uses Us

AI, Living Labor, and the Fixed Capital of Language

Artificial intelligence is usually sold as a service. You pay a subscription and receive answers, images, texts, code, translations, diagnoses, plans. The scene appears simple: a user facing a tool. But this contract hides a stranger relation. The user is not only a consumer. The user is also working. More than that, the user becomes raw material.

AI should not be understood merely as assistive technology, but as a new form of machinery. In the industrial factory, machines absorbed force, gesture, skill, and bodily time. In AI, the machine absorbs language, memory, style, archives, judgment, imagination, doubt, correction. If the modern factory was the factory of the body, AI is the factory of language.

Marx had already seen this inversion in the “Fragment on Machines”: machinery ceases to be the worker’s instrument and becomes an automatic system in which science operates as an alien power. The worker no longer commands the process; he is integrated into it as a conscious organ of a machinery that exceeds him. AI shifts this inversion into a more intimate zone. It does not mechanize only physical force or repetitive calculation, but symbolic operations that once seemed inseparable from subjectivity itself.

AI turns social intellect into private fixed capital. Everything that once circulated as common knowledge — books, images, articles, forums, archives, conversations, artworks, code, styles, public debates — is absorbed, processed, and reorganized as proprietary infrastructure. Accumulated social knowledge becomes a model, a platform, an API, an interface, a subscription.

But the operation does not end with the appropriation of the digital past. It continues with every use. Every prompt adds context. Every correction adjusts the machine. Every uploaded file expands its field of action. Every interaction teaches it something about how we ask, how we hesitate, how we want an answer to arrive. The user pays to access the machine, while also working for it.

This is why the classical figure of the consumer no longer holds. The AI user occupies three positions at once: client, worker, and raw material. Client, because he pays. Worker, because his interactions produce value. Raw material, because his language, documents, style, problems, and hesitations become processable inputs.

This is what separates AI from an ordinary tool. A hammer does not learn from the carpenter. A typewriter does not extract patterns from the writer. A library does not become more powerful because someone consults a book. AI, by contrast, turns use into training, questions into data, correction into optimization, and cognitive intimacy into resource.

Cognitive capitalism had already named the capture of knowledge, cooperation, communication, creativity, and affect. But AI introduces a harsher twist: it does not merely exploit living social intelligence; it fixes that intelligence into a machine and sells it back as a personalized service. Expropriation returns in the gentle form of assistance.

That return matters. AI does not appear as dispossession, but as help. Not as factory, but as interface. Not as boss, but as assistant. Its violence lies precisely in this softness. Domination no longer needs to appear as command; it can appear as efficiency.

Productivity is the promise: work faster, produce more versions, synthesize more information, answer more messages, design more alternatives, write more texts, program more functions, decide with less friction. But this productivity does not necessarily liberate the user. Very often, it raises the standard of performance. What was once exceptional becomes the baseline. If a machine makes it possible to produce ten proposals in an hour, soon ten proposals in an hour becomes the norm.

AI does not only increase productivity. It redefines productivity itself. It does not only accelerate tasks. It turns slowness into a defect. It does not only assist thought. It makes every non-optimized delay look suspicious. The result is not simply less work, but a new regime of comparison between human capacity and machinic performance.

This is where obsolescence enters. But it is not only jobs that become obsolete. A historical form of competence becomes obsolete: slow reading, the accumulation of an archive, the formation of judgment, memory, comparison, error, writing through internal friction, sustaining a difficulty without immediately resolving it. The victim is not only employment. It is formation itself.

Cognitive reduction is not immediate stupidity. It is the loss of friction. The machine smooths the passage between problem and solution, doubt and answer, desire and form. But this smoothing has a cost: the more we delegate, the less we exercise the uncomfortable zone where thought becomes our own. AI does not eliminate the subject. It preserves the subject as functional user, dependent operator, and corrector of an intelligence that does not belong to him.

This is why the political question is not only who owns the models or who controls the data. The question is what kind of subject is produced by a culture in which language, memory, and imagination are reorganized as automated service. Who can still afford to think slowly? Who can sustain a work, an archive, a research process, a sentence that is not immediately productive? Who remains captured in an assisted, accelerated, obedient cognition?

Artificial intelligence does not simply inaugurate a new tool. It inaugurates a new relation between knowledge, labor, and intimacy. It privatizes the general intellect, captures individual uncertainty, and sells the reduction of cognitive autonomy as personalized assistance.

The machine does not think in our place. It does something more effective: it accustoms us to thinking through it.