Intelligence as a Resource

As artificial intelligence tends to become a commodity accessible to all, in the manner of electricity or running water, it is not our intelligence that disappears: it is the way in which we exercise it that shifts. No longer towards the search for answers, but towards the formulation of questions, an activity that draws on what is most singular in us as human beings. An anthropological reflection on this displacement, and on what it changes in our practices.

A Transformation Rooted in a Long History

The emergence of generative artificial intelligence provokes reactions that oscillate between enthusiasm and anxiety, between fascination and rejection. These reactions are understandable, but they risk obscuring what is actually happening. To grasp the nature of this transformation, it is useful to situate it within a longer historical perspective, one that traces the moments when a fundamental human capacity changes its status, moving from a scarce and difficult resource to an abundant and accessible one.

The history of technology is marked by such tipping points. Memory, long carried by individuals, gradually externalised into writing, then into print, then into digital databases. Muscular force externalised into steam engines, then into internal combustion engines, then into electricity. Each time, the change was perceived as a threat against something essential. Each time, that threat proved to be, in reality, a displacement: human capacities did not disappear, they repositioned themselves.

This displacement is precisely what Bernard Stiegler analyses in Technics and Time (1994). Stiegler shows that our cognitive capacities have never been constructed in a purely interior and autonomous way: they have always stood in a constitutive relationship with external technical supports. Writing transformed our memory. Print transformed our relationship to knowledge. Algorithmic computation transformed our management of complex data. These successive externalisations did not impoverish thought: they modified what it became possible to think, and how. Artificial intelligence belongs to this history, but with one crucial novelty: for the first time, it is the capacity for reasoning itself, and no longer merely its memory or transmission, that is being externalised on a large scale.

Artificial Intelligence as a Commodity

There is a precise economic concept to describe this change of status: that of commodity. A resource becomes a commodity when it ceases to be a differential advantage and becomes a basic infrastructure, indifferently available, whose cost tends towards zero. This is true of electricity, running water, and telecommunications networks. Nicholas Carr, in his celebrated article “IT Doesn’t Matter” (Harvard Business Review, 2003), and then in his book The Big Switch (2008), was one of the first to analyse this process of commoditisation for computing: as processing power becomes accessible to everyone at the same price, it loses its differentiating character and becomes infrastructure, in much the same way as electricity did at the turn of the XXth century.

This analogy with electricity lies at the heart of the thesis formulated as early as 2017 by Andrew Ng, one of the most influential researchers in the field of machine learning: “AI is the new electricity.” AI, he argued, is poised to transform every sector of human activity exactly as electricity did at the turn of the twentieth century, not by replacing human activities, but by making them accessible in new ways, at other scales, with other possibilities. The developments of the 2020s have borne out this intuition. Access to capacities for reasoning, analysis, synthesis, and content generation, which until recently required rare skills and considerable investment, is becoming as ordinary as plugging a device into an electrical socket.

But the comparison with electricity has an anthropological reach that goes beyond purely economic concerns, and it is on this level that we must pause. When electricity became an accessible infrastructure, it did not make humans less capable of physical effort. It displaced the question of that effort. One no longer thought about producing light or heat: one thought about what one did with them. The space of possible projects was transformed. Entirely new ways of life, social organisations, and industrial architectures became conceivable that had not existed in the field of the previously possible.

This is precisely what the commoditisation of artificial intelligence is now producing, but on the cognitive plane. The space of what it is possible to want to think, to undertake intellectually, to construct as a project of knowledge, is expanding in a way that has no precedent in human history.

What the Naturalist Tradition Made of Thought

To grasp the depth of this displacement, we must return to the way our cultural tradition has conceived the activity of thinking. Philippe Descola, in Beyond Nature and Culture (2005), shows that naturalist ontology, the framework that structures modern Western thought, rests on a radical distinction between two registers: on one side, physicality, the body, matter; on the other, interiority, soul, consciousness. Within this framework, thinking is an essentially interior activity. Thought is what happens “within us,” sheltered from the external world, in a sovereign space that nothing should be able to inhabit in our place.

Descola shows that this conception is far from universal. In the animist ontologies he analyses, the boundaries of interiority are porous and relational: to think is already to be in correspondence with external entities, spirits, ancestors, forces of the living world. Tim Ingold, in Being Alive (2011), extends this intuition: for him, thought does not precede action, it takes place in the correspondence between the human being and their environment. To think is to follow the world, to respond to it, to be engaged with it. Intelligence is never a purely interior property.

This anthropological detour helps us understand that the resistance we feel towards artificial intelligence, this sense that it encroaches on something that properly belongs to us, stems largely from this very particular conception of thought as a sovereign and interior activity. Yet this conception is culturally constructed, and it has already been put to the test by every major cognitive technique. Writing itself was perceived, from its very origins, as a threat to living memory and authentic thought. Plato, in the Phaedrus, has Socrates criticise writing on papyrus scrolls as a deceptive substitute for real thought: what is fixed on papyrus cannot answer questions, does not know to whom it speaks, loses the liveliness of oral exchange. The same anxiety accompanied the appearance of the codex, the book as we know it, then of print, then of the Internet. The question is therefore not a new one. What is new is its amplitude.

The Displacement Towards the Question

What the commoditisation of artificial intelligence makes visible is a profound redistribution of what, in our cognitive activity, requires our own effort and what can be delegated. This redistribution is not new in principle, it has accompanied every major technical transformation. But it has now reached a level of generality that touches the very nature of intellectual work, and calls for reflection on what it means to think.

Hannah Arendt, in The Life of the Mind (1978), proposes a distinction whose relevance proves particularly illuminating in this context: that between thinking and knowing. Knowing is to seek answers to defined questions, to accumulate certainties, to produce stabilised knowledge. Thinking is an activity of a different nature: it sustains itself within the question rather than seeking to resolve it, it explores the depth of what remains open, it resists the closure of meaning. For Arendt, it is precisely this capacity for thinking, and not merely the capacity for knowing, that constitutes our dignity as reflective beings.

This distinction finds an echo in the hermeneutical philosophy of Hans-Georg Gadamer. In Truth and Method (1960), Gadamer defends a thesis that appears paradoxical but proves foundational: the question has a logical priority over the answer. It is not the answer that gives meaning to the question: it is the question that opens the horizon within which an answer can have meaning. “One cannot know what a proposition is,” he writes, “if one does not know the question to which it responds.” To formulate a question is therefore already to accomplish an intellectual act of the first order: to delimit a field of the thinkable, to orient a search, to decide what is worth exploring. This act presupposes a point of view, an experience, an engagement in the world, qualities that the machine does not possess.

This is also what Paulo Freire had understood, from a very different horizon, in Pedagogy of the Oppressed (1968). Freire opposed what he called “banking education”, which deposits answers in minds conceived as empty vessels, in favour of education through problematisation. To learn, for Freire, is to learn to formulate questions about the world in which one lives. It is an act that always engages a concrete situation, a lived experience, an aspiration. No machine can accomplish this act in anyone’s place, because it is inseparable from a presence in the world.

Artificial intelligence is taking over a growing share of what we used to call “knowing”: finding information, synthesising it, giving it form, drawing consequences from it according to explicit rules. This movement of delegation frees up, in principle, a more precious and more difficult capacity: that of asking the questions worth asking. Formulating a good question is not a neutral operation. It is an act that engages lived experience, singular curiosity, judgement about what merits exploration, the values that orient inquiry, and even the capacity to sustain the uncertainty that the question introduces before an answer comes to close it.

It is in this sense that the transformation underway can be read as an invitation to deepen what is most properly our own. Not to defend a threatened territory, but to inhabit more fully a territory we occupied only partially: that of the question, of intention, of meaning, of the sensory.

The Mobility Analogy: Expanding the Space of the Possible

The comparison with the revolutions in mobility is useful for understanding the nature of this enlargement. The appearance of the internal combustion engine, then of the railway, then of aviation, did not simply make movement faster. It transformed the way human beings conceived their projects. People began to imagine ways of life, relationships, and forms of organisation that were simply unthinkable before these technologies existed. The train did not render legs useless. It made possible a space of new intentions.

Artificial intelligence operates an analogous displacement, but in the order of intellectual possibility. Problems whose complexity made them inaccessible to any single person, syntheses that would have required years of research, connections between bodies of work that no one could have read in their entirety in a lifetime, all of this becomes available on demand. This is not merely an acceleration of what was already being done. It is the opening of a space of projects that did not exist in the field of the thinkable.

Recent data from scientific research illustrates this enlargement concretely, while also revealing its tensions. A study published in Science in December 2024 by Yian Yin and colleagues at Cornell University analysed nearly 2.1 million preprints deposited between 2018 and 2024 on the three main international preprint platforms: arXiv (physical sciences, mathematics, computer science), bioRxiv (life sciences), and SSRN (social sciences and humanities). The results show that researchers using large language models to write their work significantly increased their output: +59.8% in the social sciences and humanities, +52.9% in biology, +36% in the physical sciences. A further study, published in Science Advances, estimates that at least 13.5% of all scientific abstracts published in 2024 were written with AI assistance. The Cornell study also notes a notable qualitative effect: researchers using AI draw on a more diverse and more recent bibliography, establishing connections with work they would not have identified through traditional search methods. But the same data calls for caution: papers with a heavily AI-shaped writing style were, on average, less often accepted by peer-reviewed journals than their high-quality human-authored counterparts. This paradox is precisely what this article seeks to name: the power of the tool does not determine the value of the work. What makes the difference is the quality of the questions researchers put to it, their capacity to formulate problems that are worth pursuing.

For this enlargement, the same warning applies as for the revolutions in mobility: it does not automatically produce improvement. The car also produced traffic jams, dependence on hydrocarbons, the destruction of city centres, urban sprawl, the worsening of pollution, and immense public health problems. Artificial intelligence will produce its own forms of alienation if we confine ourselves to a purely consumptive relationship with its capacities, without developing the reflexivity that allows us to direct its use.

The issue is not the technology in itself, but the quality of the questions we put to it and the nature of the projects we entrust to it. And this quality is not given: it is constructed. It requires that we have developed, through our own work on ourselves and on our relationship to the world, the capacity to formulate questions that arise from real experience, from grounded curiosity, from an intention that can only come from us. This is precisely what the commoditisation of artificial intelligence brings to the fore: no longer the mastery of answers, but the cultivation of questioning.

What This Changes in Practice: Concrete Orientations

The anthropological mutation I am describing is not a prophecy about the future. It describes what is already happening, right now, in the everyday practices of work, creation, and learning. Drawing practical consequences from it calls for a few clarifications.

The first is to take seriously the distinction, which I have developed elsewhere, between tasks, occupations, jobs, and work in its deeper sense, the sense that shares its root with travel, with journey, with interior transformation. Artificial intelligence takes over a growing number of tasks, reconfigures occupations, transforms jobs. But it cannot perform work in the sense of an encounter with otherness, the creation of meaning in a particular context, an embodied relationship. This register remains entirely human, and it is towards it that the commoditisation of artificial intelligence invites us to shift.

The second clarification concerns education and learning. Delegating to AI the cognitive operations that require effort carries a real risk: short-circuiting processes of formation that necessarily pass through difficulty, error, and return. The resistance one feels when searching for an answer oneself is not an obstacle to be eliminated: it is often the very condition of understanding. One must therefore distinguish between using AI as an amplifier of capacities already developed and using it as a substitute for capacities not yet developed. The first use is valuable; the second is impoverished.

The third clarification touches on the question of inequalities. The history of commoditisation shows that resources that become infrastructure do not thereby become universally accessible. Electricity remained for decades a privilege of cities, before reaching rural areas and developing countries. Artificial intelligence is tending to become a commodity for those who have access to it, but this access remains very unequally distributed. The anthropological mutation I am describing does not unfold uniformly: it first deepens the gaps between those who can put questions to powerful systems and those who cannot.

Correspondence Rather Than Consumption

There remains a question that these practical orientations do not entirely resolve, and that it would be dishonest to avoid. The commoditisation of artificial intelligence does not automatically produce an elevation in the quality of the questions asked. It can also produce, and is already producing, a laziness of thought: the immediate resort to the available answer, the premature closure of inquiry, satisfaction with the first plausible result. This is a risk symmetrical to that of non-use: no longer resisting the tool out of fear, but abandoning oneself to it out of comfort. To name this risk precisely, one must distinguish between two possible relationships to artificial intelligence, which Tim Ingold, in Correspondences (2021), would help us articulate in his own way.

Ingold opposes two modes of being in the world: transport and wayfaring, the charted crossing and the act of following a path. In transport, one moves from one point to another along a predefined route; one consumes space without truly inhabiting it. In wayfaring, one traces a path by walking, one responds to the solicitations of the terrain, one improvises according to what one encounters. It is not efficiency that wayfaring lacks: it is a different quality of attention, a different way of being present to what is happening. The person who walks learns something that the passenger does not.

This distinction applies exactly to our relationship with artificial intelligence. One can use it in the mode of transport: pose a conventional question, receive a standard answer, consume the information without traversing it. Or one can use it in the mode of wayfaring: engage an exchange that follows thought wherever it goes, that rebounds off what the answer unexpectedly reveals, that uses the results as material for new questions. In the first case, AI relieves us of an effort. In the second, it amplifies an effort that remains entirely our own.

What Ingold calls correspondence, in the double sense of an exchange of letters and of co-response, of responding-together, designates precisely this second mode. To correspond with AI is not to consult it as one consults an index: it is to engage with it in a movement of thought in which one brings what one is, one’s experience, one’s curiosity, one’s point of view, and from which one receives in return what one would not have formulated alone. The condition of this correspondence is exactly the same as that of any productive relationship: to be present, attentive, disposed to be surprised.

Sophie Nordmann, in The Vocation of the Philosopher (2025), formulates with precision what makes this presence both possible and necessary: “What is proper to human thought is a hole, a gap. When this breach closes, all is lost.” The breach she speaks of is the space of questioning that does not close, of curiosity that resists satisfaction with the first result, of thought that keeps the question open beyond the answer. It is this nothing that changes everything, she writes. It is there that the irreducible specificity of what we do when we think resides, in the sense that Arendt gave to that word, and in the sense to which the commoditisation of AI now invites us to attend.

The anthropological mutation of our time therefore cannot be resolved in the mere technical mastery of the tool. It demands a culture of questioning, a vigilance towards one’s own thought, a resistance to the closure of meaning, an attention to those moments when an answer opens more than it closes. It demands, more profoundly, that we remain beings who wayfar rather than passengers. Not out of nostalgia for a vanished effort, but because it is in this wayfaring, in this active, improvising, detour-sensitive attention, that the capacity to ask the questions worth asking is built. And it is precisely this capacity that allows artificial intelligence, as a commodity, to become something other than an infrastructure of comfort: a partner in thought, on the condition that we are sufficiently present to put to it questions that deserve an answer.