AI and the New Machinery of Dispossession

 AI and the New Machinery of Dispossession

There is a habit, when a new technology arrives, of treating it as though it were self-contained. A machine appears, or a system, or now a model, and we speak of it as if it had descended from nowhere, immaculate, severed from history. We say: here is artificial intelligence. What will it do? Will it write our memos, drive our cars, diagnose our illnesses, perhaps compose our poems badly and our legal briefs passably? And because the machine dazzles, because it seems to speak, and because it answers in a tone of synthetic confidence, we imagine that the story begins there, with the glowing surface.

But it does not.

It begins much earlier, in an older pattern of human arrangement. It begins with the persistent genius of modern capitalism for finding new things to enclose, new spaces to reorganize, new habits to turn into property, and new forms of cooperation to convert into private gain. Artificial intelligence, in that sense, is not merely a technology. It is an episode in a much older drama.

David Harvey, the geographer, gave us a language for some of this. He spoke of accumulation by dispossession, which is to say the process by which wealth is not created only by making new things, but by taking hold of what was once shared, public, customary, or only loosely owned, and bringing it under control. Land, water, public goods, social practices, knowledge: all can be drawn into the machine.

Seen this way, AI is not simply a clever engine for answering questions. It is also a system for appropriating the products of collective life. Language itself is social. Images are social. The styles of artists, the habits of coders, the judgments of teachers, the accumulated stock of public writing, public argument, public memory, all of this becomes raw material. The machine is trained on what multitudes have made, but the resulting power is frequently concentrated in a small number of firms. What was diffuse becomes centralized. What was shared becomes asset. What was lived becomes data.

That is dispossession.

And yet dispossession alone does not explain the peculiar physicality of the AI boom, because AI is often spoken of as if it were weightless, as if it floated somewhere above the earth in “the cloud,” which is one of the more successful acts of industrial misdirection in modern language. A cloud suggests vapor, softness, nothing you can touch. But AI has weight. It sits on land. It drinks water. It consumes electricity on a scale that would have startled the engineers of the postwar world. It requires substations, transmission lines, concrete pads, warehouses of servers, chip fabrication plants, cooling systems, mines, logistics corridors.

So now another of Harvey’s ideas becomes useful: the spatial fix. Capitalism, when it runs into difficulty, when profits are squeezed or growth slows, has a way of seeking relief by moving outward, by remaking space, by pouring money into infrastructure, into new built environments, into geographical rearrangements that absorb surplus capital and open a temporary avenue of expansion. Railroads did this. Suburbs did this. Container ports did this. Data centers may be doing it now.

And so the story extends. A model trained on the language of millions requires a data center. The data center requires land, tax abatements, power deals, water access, transmission upgrades. Rural counties are courted. Utilities are restructured. Public resources are redirected toward the needs of private computation. The machine that seemed to be only a sentence-completer turns out to be a regional planning event. It turns out to be a struggle over who gets the electricity, who pays for the wires, whose water is drawn down, and which communities are told that this is progress.

That is the spatial fix.

But perhaps the most important part of the story, because it is the part that people feel in their bodies, is labor. There is a melodramatic version of the AI future in which the machine arrives one Tuesday and everyone is fired by Friday. That is not, for the moment, the most plausible version. What is more plausible, and in some ways more insidious, is something quieter. Hiring slows. Vacancies go unfilled. Junior workers are not replaced. Experienced workers find that parts of their judgment, their style, their accumulated craft are being extracted, formalized, and fed back to them as supervision. The machine does not instantly eliminate the worker; it first disciplines the worker.

This is a very old pattern. Technology under capitalism is rarely introduced simply to liberate human beings from drudgery. It is introduced inside a system of power, and therefore it tends to serve power first. If demand is weak, firms use new tools not to shower society with abundance, but to defend margins, to reduce labor costs, to standardize performance, to increase managerial visibility, and to weaken bargaining power. The benefit to the firm is immediate. The benefit to society is conditional.

Here we arrive at the contradiction. AI is sold as a productivity miracle. And perhaps, in time, it will be one. But if the gains are realized chiefly through cost cutting, through labor suppression, through the concentration of intellectual property and infrastructure control, then the very force that raises efficiency can also depress mass purchasing power. Individual firms become leaner while the system as a whole becomes more brittle. The machine can produce more, but the public, made anxious or precarious, may be less able to buy. This is one of capitalism’s oldest tricks: solving the problem for the firm while deepening the problem for the society.

That is why the present moment feels so strange. We are told that a great transformation is underway, and yet for many people daily life feels not triumphant but pinched. The economy softens. Consumers hesitate. Employers grow cautious. The future is loudly advertised, but the present feels like a waiting room. This is not a contradiction at all, not really. It is what early technological transition often looks like. The infrastructure rises first. The hype arrives first. The speculative money arrives first. The broad social benefit, if it comes, arrives later, and only after institutions, labor markets, and political arrangements have been forced into a new shape.

So the real question is not whether AI is intelligent, or whether it can pass this or that exam, or whether it can produce an image in the style of a dead painter. The real question is: what social order is being built around it? Who owns the models? Who controls the infrastructure? Who captures the gains? Who absorbs the losses? What forms of work are enhanced, and what forms are hollowed out? Which towns get the tax base, and which get the water bill? Which kinds of knowledge are honored, and which are quietly strip-mined?

If one asks only what the machine can do, one sees the gadget. If one asks what world the machine requires, one begins to see the system.

And that, perhaps, is the necessary correction. AI is not merely the next device in a sequence of devices. It is a new chapter in the long history of enclosure, labor discipline, infrastructural expansion, and uneven development. It may yet generate extraordinary benefits. It may cure diseases, accelerate discovery, lower certain costs, widen certain forms of access. But those outcomes are not guaranteed by the machine itself. They depend on politics, on institutions, on ownership, on whether the immense social intelligence from which these systems are built is allowed to return as a public good, or merely comes back to us as rent.

The machine speaks in the language of the commons. The question is whether it will serve the commons, or dispossess it.