The Enclosure of the Mind by Scott Timcke
This guest blog was authored by a DISC Virtual Visiting Fellow Scott Timcke.
We have become accustomed to describing AI as a revolution in knowledge. Venture capitalists and technology executives speak incessantly of intelligence, creativity, expertise, and problem-solving. Nobody talks much about ownership. In this discourse, intelligence appears as a universal resource becoming more widely available. The more consequential development, from a Marxian political economy standpoint, lies in a set of technical and institutional arrangements through which knowledge can be detached from the individuals who possess it and reorganised as infrastructure subject to ownership, investment, and rent extraction. Capital is gaining access to skill without granting labour access to capital.
This is not a claim about conscious intent on the part of particular firms or investors. It is a description of the direction in which investment, ownership, and technological development are converging. Capital has always sought to reduce its dependence upon labour. This is among the oldest observations in economics. Machinery mattered less for the productivity gains it delivered than for how it altered social relations within production. The factory system reduced reliance upon artisanal knowledge; Taylorism reduced reliance upon worker discretion; automation reduced reliance upon physical labour. AI extends the same logic into occupations that had previously enjoyed a degree of insulation from mechanisation because their value rested upon specialised forms of judgement, interpretation, and expertise.
For much of the twentieth century, advanced capitalist economies generated expanding strata of professional workers whose market position rested upon credentials, training, and forms of knowledge that could not easily be codified. Lawyers, accountants, consultants, engineers, architects, software developers, and analysts sold their labour power, yet retained leverage because organisations could not readily separate expertise from the people who carried it. Knowledge remained stubbornly embodied.
Contemporary AI systems do not so much eliminate expertise as render it portable, reproducible, and available independently of its original bearers. Whether these systems ultimately fulfil the more extravagant promises made on their behalf is beside the point. Even partial success alters the balance of dependence. A corporation that once required a substantial cadre of analysts may increasingly require a model and a smaller supervisory layer. A legal practice may discover that routine work can be reorganised around computational systems overseen by fewer professionals. Expertise itself may become available in a form that can be rented instead of hired.
The history of capitalism is, in one telling, a history of successive commodifications. Land became property. Labour became wage labour. Information became intellectual property. Now, access to organised cognitive capacity is assuming a similar form. The token illuminates this transformation. As firms move towards token-based pricing models, intelligence acquires a measurable and exchangeable form. One purchases neither software in the traditional sense nor even information, but quantified access to compute. The appearance of token-price indices is worth noting precisely because benchmarking generally follows commodification. Once markets begin comparing, indexing, and hedging a resource, that resource has usually crossed an important threshold in economic life.
That shift in the form of expertise goes some way toward explaining the extraordinary scale of investment currently flowing into AI. Much commentary draws comparisons with the dot-com boom, but the analogy is not a perfect fit. The internet boom was characterised by speculative expectations that ran ahead of material realities. AI, by contrast, is anchored in immense and growing physical infrastructure. Every increment in computational capability requires semiconductors, advanced packaging, memory, fibre-optic networks, data centres, cooling systems, and vast quantities of electricity. The constraints are physical, not just financial, and investment has propagated outward through an entire ecosystem of supporting industries, from semiconductor fabrication and energy generation to construction, logistics, and grid expansion. The physical footprint of AI looks less like a software sector and more like heavy industry.
Ownership has concentrated accordingly. Training frontier models requires resources available only to a small number of firms. Compute now functions as fixed capital, and access to expertise is mediated through infrastructures controlled by a narrow group of corporations. The crucial assets go well beyond algorithms. Data centres, chip supply chains, energy contracts, and capital reserves are the real stakes.
Current pricing obscures these underlying relations. Token prices remain remarkably low relative to the capital expenditure required to sustain the industry, while the leading firms continue to absorb enormous losses while expanding capacity. The economic strategy is an old one. Railways, telecommunications networks, and digital platforms all passed through periods in which access expanded rapidly under conditions of heavy investment and uncertain profitability, services subsidised to accelerate adoption and entrench dependence, markets captured before they were monetised. Clarity about what had happened often arrived late, when infrastructures that initially appeared open had quietly acquired the characteristics of essential utilities while remaining privately controlled.
The models themselves are trained upon immense accumulations of human activity, like books, articles, software repositories, technical documentation, public discussion, and countless other forms of collective intellectual labour. Yet access to the resulting systems is organised through proprietary platforms and priced through proprietary mechanisms. The inputs are collective; the infrastructure is private.
The most enduring questions raised by AI may have little to do with cognition. They concern ownership, access, and distribution. It is the classic question of control rights. Who controls the systems through which expertise is delivered, who captures the returns generated by automated judgement, and under what conditions knowledge created collectively becomes the basis of private accumulation.
Most AI ethics debate clusters around bias, transparency, safety, and alignment, with the model or its output treated as the primary site of concern. These questions matter, but they draw attention to surface-level properties of systems whose deeper significance lies in their ownership structures. Once AI is treated as part of a broader reconfiguration of ownership over cognitive infrastructure, the scope of ethical analysis has to widen beyond the behaviour of models to include the institutional and political economy within which those models are produced and deployed. Once this step is taken, fairness and explainability do not look not like technical properties to be optimised within a fixed system, but downstream expressions of prior decisions about ownership, investment, and access. The problem of aligning machine behaviour with human values turns out to be inseparable from the problem of aligning institutional arrangements with democratic control over the means through which intelligence is produced and distributed.
The token is, at one level, a technical accounting unit. At another, it is a small indicator of a larger transformation of converting intellectual capacities into metered services delivered through privately owned infrastructures. Behind the token lies the server rack, behind the server rack the data centre, behind the data centre the electrical grid, financing arrangements, land acquisitions, and industrial supply chains. The history of AI may ultimately be written less as a history of machine learning transformers than as a history of the transformation of how ownership around cognition.
