AI and the Means of Intellectual Production
AI isn’t so much a break with past history as a continuation of capitalism’s accelerating tendencies—with humanity’s collective data as its training set, and with millions of jobs at risk.
Silicon Valley’s technologists have vacuumed up humanity’s near-total cultural output to build platforms with tremendous potential economic and ideological power. Generative AI has become the subject of enormous hype and staggering, trillion-dollar investments. So what can we learn about AI by reading it through the lens of a nineteenth-century thinker like Karl Marx?
What’s striking about AI, from a Marx-inflected perspective, is how little is fundamentally new about this technology in political-economic terms. In the Grundrisse, Marx describes how capitalism generates ever more advanced machines and scientific and technical knowledge, which become increasingly important factors in the production of wealth. In Capital, Marx shows that machines, from the capitalist’s standpoint, serve as a means of cheapening commodities and intensifying the working day, raising the level of surplus-value extracted from workers.
From this vantage point, AI is less a revolutionary new force than the latest iteration of a centuries-long process of acceleration and intensification of economic production. When a Meta executive tells employees to “go 5x faster” using AI tools, this is not fundamentally different from what Henry Ford aimed to accomplish with the assembly line in 1913. As the technology historian David Edgerton has shown, we have repeatedly assumed that new technologies will be truly transformative. AI’s cheerleaders should not mislead us into treating AI as a fundamental rupture with history, but rather as part of capitalism’s ongoing intensification of the labor process—along with, it must be said, an attempt to justify massive AI company valuations amid still limited profitability. There are legitimate concerns about an AI bubble—that is, overinvestment in an as-yet largely unprofitable technology.
Four Marx-derived concepts, I claim, can help us think critically about what’s happening around us.
Primitive accumulation
In Capital, Marx describes how England’s rural commons were enclosed and privatized, with public land appropriated from the people without compensation serving as the very precondition for capitalism’s rise. Something analogous is happening with AI. In 2025, it was alleged that Meta had systematically pirated over 81 terabytes of copyrighted e-books to train its language models. Meanwhile, freely available resources like Wikipedia and the Common Crawl dataset—hundreds of billions of web pages—were foundational to training platforms like GPT-3.
What has been appropriated approximates Marx’s general intellect—the term he uses in the Grundrisse to describe the shared stock of scientific and technical knowledge that becomes a direct force of production: the accumulated social knowledge of humanity, now in objectified form—or as Kate Crawford puts it, through a commercialized “capture of the commons.” This has been done largely without any sort of compensation to individual authors for their writings or users for their data; one approach might be to compensate society as such, standing for the very universality that has been appropriated, perhaps in the form of free model access to the public, or licensing revenues to governments, as is done with many natural resources around the world.
Alienation
Marx showed how industrial production alienated workers from their products, their labor, and ultimately from themselves; generative AI furthers this alienation in new ways. As more text, images, and video circulating in our culture becomes machine-produced, we gradually risk becoming estranged from language and “signs” themselves as well as from human-to-human communication. The author is being rewired into something like a cyborg, a human-machine hybrid whose “authenticity” can never fully be verified. Whether this is truly problematic in the long run or whether it will give rise to new forms of human/machine interfacing and human sociability remains to be seen, but all of this has the property of a giant social experiment—importantly, one conducted without any sort of ethical research protocols.
Cognitive offloading—deskilling through repeated reliance on AI tools—is also a risk. Studies suggest AI may create a kind of “cognitive debt.” A Lancet study found that doctors using AI tools to detect precancerous growths became significantly worse at finding them on their own after just three months. MIT researchers have claimed in a preprint study that students using ChatGPT showed markedly less brain activity than those limited to conventional search engines while writing essays. As the “general intellect” is increasingly absorbed into black-box AI models, the technology risks becoming a replacement of human cognitive capacity.
Ideology
When AI speaks, who is really speaking? Platform defenders claim these are neutral tools, merely recombining earlier human utterances. But precisely because of this, the models will inevitably reproduce dominant thought patterns and existing inequalities—what writers like Cathy O’Neil and Safiya Umoja Noble have documented as algorithmic bias. A recent Stanford study showed that AI-powered applicant screening tools are susceptible to racial bias: in a study of 3.4 million applicants, the authors estimated that an additional 40,000 Black and Asian applicants would have advanced to the next stage if their applications had been treated as favorably as the “most-favored group (typically white applicants).”
We also see signs that the regnant worldviews of Silicon Valley’s tech industry will shape AI’s “voice”—I am probably not the only person to have noticed that some LLMs seem to speak in the unique parlance of Silicon Valley-speak, from “belt-and-suspenders” approaches to “load-bearing,” “legible,” “spines” and other vaguely techy, managerialist phrases. The form of AI speech encodes the social context of those who built it. China’s DeepSeek refuses to engage with questions about Tiananmen; Elon Musk’s Grok began spouting antisemitic statements and calling itself “MechaHitler.”
As Marx and Engels wrote, “The ruling ideas of each age have ever been the ideas of its ruling class.” When billions of people increasingly filter their information through a mere handful of AI platforms, the ideological stakes are enormous.
Platform power
Marx observed that capital tends toward concentration until finally the “expropriators are expropriated,” as he writes in Capital. Nick Srnicek has updated this insight for the digital age, studying how platforms become monopolistic intermediaries controlling vast flows of data, labor, and consumption. The AI revolution intensifies this dynamic: a handful of giants—Alphabet, OpenAI, Microsoft, Anthropic—now control the dominant AI platforms, acquiring in the process a new and deeper, more insidious form of informational power than the monopolists of the previous dotcom era.
Some Marx-influenced thinkers, like Yanis Varoufakis, have gone so far as to call this “technofeudalism”—a stage where surplus is extracted not through free wage labor but through infrastructural control over digital ecosystems in a manner resembling precapitalist feudalism; it is a term I am not enthusiastic about because its anachronism if anything obscures the deepening and intensification of contemporary tendencies under capitalism, which are not “throwbacks” to previous stages in political-economic history.
What matters is that trillion-dollar AI companies will wield ever-widening forms of informational and infrastructural power through their control of the platforms through which millions of people increasingly learn and labor.
Toward a public AI?
The political battle over AI is only just getting started. If productivity gains do materialize, the question is: who reaps the benefits, and who bears the costs? Just as natural resource extraction is subject to taxation, the “extraction” of cultural resources by AI companies ought to be subject to robust public regulation and taxation. Without regulation, we risk an “anything goes” situation; without strong redistributive mechanisms, AI will continue to forge new tech oligarchs, only now surrounded by millions of potentially redundant, unemployable workers. AI companies should be made to absorb some of these costs.
It is, however, entirely possible to imagine a public, non-commercial AI—transparent, accountable, and modeled on experiments like Wikipedia or the original vision of the World Wide Web. This will require significant public investments. Without such public, nonprofit AI systems, we risk ceding ever more economic and political power to a small class of tech barons. Increasingly, future political struggles look set not to be over “AI or no AI,” but over the weights, the training data, the system prompts, and the rightful allocation of compute. The struggle over AI is, fundamentally, a struggle over who controls the means of intellectual production, and it has only just begun.
A longer version of this argument appears in Norwegian in Agora (no. 1–2, 2026).