# When intelligence stops being scarce: work, technique and the human question after AI **Program:** Technology, Production & Society **Code:** MT-TSE-2026-10-05-intelligence-abundance-human-question **Edition:** 5 October 2026 **Information cutoff:** 5 October 2026 For almost all of economic history, applied intelligence was inseparable from people. To obtain more calculation, more writing, more design, more interpretation or more coordination, someone had to be trained, hired and organized. Even when machines replaced physical force, command, language, judgment and supervision remained attached to human bodies, careers, professions and institutions. Artificial intelligence changes that relationship without necessarily abolishing any of those things. For the first time at industrial scale, a growing share of language, classification, synthesis, code, prediction and coordination can be reproduced as a technical service. The social novelty is not that a machine “thinks like a person.” It is that activities previously constrained by the amount of qualified human time available can increasingly be reproduced at much lower marginal cost. At first glance, this looks like a productivity problem. It is deeper than that. Work is also a structure of recognition, time discipline, identity formation, income distribution, knowledge transmission and social participation. If a society learns to produce more with less human time, it does not automatically receive more freedom. It may receive unemployment or leisure; concentration or abundance; autonomy or dependence; greater creativity or intensified surveillance. Technology opens possibilities. Institutions largely determine which possibilities become normal. The decisive question, then, is not whether AI will be “good” or “bad.” It is this: **what kind of human being and what kind of society become possible when operational intelligence becomes less scarce while ownership, power, recognition and meaning remain scarce?** Versions of this question appeared long before generative models. Hannah Arendt separated labor, work and action. Karl Polanyi showed how markets can reorganize social relations when labor and nature are treated as commodities. Karl Marx imagined production in which social knowledge becomes embodied in machinery. Norbert Wiener warned that automation is also a problem of communication, control and purpose. Jacques Ellul examined the expansion of efficiency as an organizing principle. Lewis Mumford argued that large technical systems are social systems as well. Ivan Illich asked whether tools expand autonomy or produce dependence. Richard Sennett studied what flexible capitalism does to continuity, character and biography. John Maynard Keynes imagined a society in which the economic problem might cease to occupy the center of life. These writers disagree profoundly with one another. That is precisely why they are useful. None provides a ready-made answer for AI. Together, however, they move the question from machine performance to the organization of life. ## Before asking what AI will do, we need to ask what we mean by work A recurring difficulty in debates about automation is treating “work” as one thing. In *The Human Condition*, Hannah Arendt offers a distinction that remains powerful: **labor**, **work** and **action**. [University of Chicago Press — The Human Condition](https://press.uchicago.edu/ucp/books/book/chicago/H/bo29137972.html) Labor refers to activities required to sustain life and therefore repeated: feeding, cleaning, producing, maintaining. Work creates a relatively durable world of objects, institutions, artifacts and forms. Action, by contrast, takes place among people and involves speech, initiative, plurality, unpredictability and public presence. Industrial economies mix these spheres. An engineer may spend part of the day doing repetitive labor, part building something durable and part acting politically inside an organization. A teacher repeats content, produces materials and also participates in relationships that cannot be reduced to information transfer. A physician classifies symptoms, records data and at the same time assumes responsibility before a concrete person. AI does not affect these three dimensions in the same way. ```diagram Labor repeatable tasks · classification · recording · routine transformation ↓ higher automation potential Work design · construction · integration · creation of durable systems ↓ strong amplification potential Action responsibility · promise · conflict · trust · public decision ↓ can be mediated by AI, but legitimacy remains a human and institutional problem ``` A society that confuses these dimensions can interpret the automation of a task as the automation of a profession, or the generation of an answer as the replacement of a relationship. The distinction also explains why “can the model do it?” is not enough. Technical ability to produce a legal text does not settle who is accountable for it. The ability to generate a diagnosis does not decide who bears the risk. The ability to synthesize political arguments does not create democratic legitimacy. The ability to write code does not choose which systems ought to exist. The more abundant operational intelligence becomes, the more visible everything becomes that was never merely intelligence. ## Polanyi: markets do not merely distribute goods; they reorganize social relations In *The Great Transformation*, Karl Polanyi described the formation of a society in which economic relations ceased to be merely embedded in social obligations and increasingly reorganized society itself. His analysis became famous in part for the idea of labor, land and money as “fictitious commodities”: elements fundamental to social life treated as if they had been produced for sale. [Beacon Press — The Great Transformation](https://www.beacon.org/The-Great-Transformation-P46.aspx) The parallel with AI does not require claiming that data or intelligence are simply new fictitious commodities. The deeper point lies in the mechanism: when a human capability can be measured, bought, sold and substituted through markets, institutions begin to reorganize around that form of calculation. For much of the twentieth century, firms bought human time in relatively legible units: hours, salaries, roles and teams. Digitization made tasks more measurable. Platform economies fragmented work into rides, deliveries, clicks, tasks and ratings. AI systems add another possibility: fragments of intellectual work can be priced in calls, tokens, actions, resolutions and outcomes. The transformation is not merely technical. It changes the accounting unit of human activity. When software is sold per user, the employee remains the economic unit of the product. When an agent is sold per action or outcome, the task itself begins to become the unit. The boundary between “buying software” and “buying labor” becomes less clear. The Polanyian risk appears when society treats this transformation as a purely technical fact. A price can measure task efficiency without measuring community stability, formation of future professionals, psychological security, trust, autonomy or public participation. An institution that cuts costs can create an externality absent from its spreadsheet: fewer entry-level jobs today can mean fewer specialists trained tomorrow; more administrative automation can free time or simply raise expected output; less need for labor can raise disposable income or weaken bargaining power. A technology’s great transformation is not contained inside the technology. It occurs when rules, markets and habits adapt around it. ## Marx: when social knowledge becomes a direct productive force In the *Grundrisse* manuscripts of 1857–1858, Marx explored an extraordinarily modern possibility: scientific and social knowledge could become embodied in machinery to such an extent that productive power would depend less directly on each person’s immediate labor. Penguin’s edition highlights the notebooks’ treatment of automation, labor, surplus value and alienation. [Penguin — Grundrisse](https://www.penguin.co.uk/books/35199/grundrisse-by-karl-marx-translated-with-a-foreword-by-martin-nicolaus/9780140445756) The relevance for AI is not that Marx somehow predicted language models. It is that he identified a problem of ownership. Knowledge is never produced by an isolated individual. Language, mathematics, science, programming, literature, administration, engineering and law are historical accumulations. Each generation inherits a cognitive infrastructure built by others. Technology converts part of that inheritance into productive capacity. When a model learns patterns from enormous corpora and then performs tasks, a portion of accumulated social knowledge reappears as a technical asset. The economic question becomes: **who owns the mechanism that turns collective knowledge into private production?** There is no simple answer. Models require investment, engineering, energy, hardware and organization. Firms assume risk and create real technology. At the same time, the cognitive raw material is inseparable from a social history far larger than any single company. The tension is structural. The more production depends on accumulated knowledge, the less intuitive it becomes that income should track individual hours of labor directly. Yet if income remains organized mainly through employment, a society can produce technical abundance and economic insecurity at the same time. This is one of the central contradictions of the AI era: **productivity can become more social while ownership remains concentrated**. ## Wiener: automation is not only substitution; it is an architecture of control Norbert Wiener, one of the founders of cybernetics, wrote *The Human Use of Human Beings* in 1950 and revised it in 1954. His problem was not simply how to build automatic machines. He was concerned with communication, feedback, control and the social consequences of systems able to respond to their environment. [Google Books — The Human Use of Human Beings](https://books.google.com/books/about/The_Human_Use_Of_Human_Beings.html?id=DydKDgAAQBAJ) That tradition is useful because it shifts attention from “the robot that replaces the human” to the **circuit** in which humans and machines participate. An automated system never exists alone. Someone chooses its function, metrics, data, right to act, tolerance for error and the point at which a decision must return to a person. ```flow Institutional objective → chosen metric → automated system → action → observed outcome → feedback → behavioral adjustment → new institutional pattern ``` Power lies as much in the design of that circuit as in the model itself. AI used to reduce bureaucracy can return autonomy to a worker. The same AI, attached to permanent monitoring and automatic targets, can make the worker more subordinate to metrics they do not control. A recommendation system can widen access to knowledge or narrow attention by optimizing only engagement. A tutor can personalize learning or turn education into a sequence of answers that removes enough cognitive effort to weaken understanding. Wiener’s question remains current: it is not enough to know whether a machine works. We must know **for what purpose, under what feedback loop and for whose benefit**. ## Ellul: the danger is not only machines, but efficiency becoming a norm Jacques Ellul used the French word *technique* more broadly than “technology.” In *The Technological Society*, first published in French in 1954, he was concerned with the spread of rationalized methods of maximum efficiency across domains of life. The International Jacques Ellul Society emphasizes that for Ellul, *technique* cannot be reduced to machines; it is a mode of organizing activity according to criteria of effectiveness. [International Jacques Ellul Society](https://ellul.org/themes/ellul-and-technique/) [Google Books — The Technological Society](https://books.google.com/books/about/The_Technological_Society.html?id=TEs0AAAAMAAJ) The diagnosis becomes more unsettling with AI because the system does not only execute procedures; it helps define, optimize and replicate procedures. The problem appears when efficiency stops being a tool and becomes an implicit moral criterion. If a professional can produce ten times more with AI, why work less instead of producing ten times more? If a school can generate personalized material instantly, why preserve slow activities? If a decision can be probabilistically classified, why tolerate slow judgment? If an agent can resolve a conversation, why preserve costly human relationships? If a process can be measured, why allow zones that remain unmeasured? None of these questions has a technical answer. Technique tends to remove what appears to be waste. But some forms of waste are constitutive of human life: conversation without an objective, learning without immediate application, redundant care, contemplation, ritual, friendship, exploratory research, trial, error and unoptimized time. A society can become more efficient and less inhabitable. ## Mumford: the most powerful machine can be an organization In *The Myth of the Machine*, Lewis Mumford argued that large systems of power do not depend only on mechanical devices. What he called the “megamachine” includes the coordination of people, authority, information and technique in structures able to execute projects at scale. The two volumes, published in 1967 and 1970, treat technology as part of a broader social formation. [Google Books — The Myth of the Machine](https://books.google.com/books/about/The_Myth_of_the_Machine.html?id=FtIyAAAAMAAJ) The idea is especially useful for understanding contemporary AI systems. The model is only one layer. Around it sit data centers, chips, electricity, cloud contracts, platforms, datasets, interfaces, safety standards, corporate policies, evaluation teams, identity systems, legal departments, labor markets and governments. The real “machine” is the combination. ```mindmap The social machine of AI - Infrastructure - chips - electricity - data centers - networks - Knowledge - data - science - language - software - Institutions - firms - universities - governments - markets - Control - identity - permissions - metrics - audit - Everyday life - work - education - consumption - communication - culture ``` This changes the debate about “controlling AI.” There is no single button capable of controlling a distributed social system. Control means governing infrastructure, incentives, ownership, standards, access, accountability and the capacity to contest decisions. The Mumfordian risk is that systems become so large and interdependent that preserving their continuity begins to justify every decision. The institution stops asking “does this serve human life?” and begins asking “what must we do to keep the system operating?” ## Illich: a tool is good when it increases the capacity to act without making the user dependent on it In *Tools for Conviviality* from 1973, Ivan Illich distinguished between tools that expand autonomous action and systems that make people dependent on specialized structures. His idea of “conviviality” was not a rejection of technology but an attempt to preserve the capacity of people and communities to use tools without being subordinated to them. [Google Books — Tools for Conviviality](https://books.google.com/books/about/Tools_for_Conviviality.html?id=EgKaPwAACAAJ) This may be one of the most productive questions for evaluating AI: **after using the tool, is the person more capable or merely more dependent?** A programmer who uses AI to explore code and ends up understanding the system better may gain autonomy. A programmer who stops understanding what they deliver may gain speed while losing capability. A student who uses a tutor for adaptive explanations may learn more. A student who outsources every difficulty may learn how to avoid the act of thinking itself. A small entrepreneur able to produce software, design and analysis with accessible tools may challenge larger firms. The same entrepreneur may become completely dependent on platforms controlling price, access, distribution and rules. A convivial tool lowers barriers without removing agency. An infrastructure of dependence offers convenience in exchange for exit capacity. That criterion is more useful than asking whether a technology is “open” or “closed” in the abstract. Open systems can still require concentrated infrastructure; proprietary systems can still create real new capability. The practical question is how much decision power remains with the user and how much is embedded in the architecture. ## Sennett: work organizes a story about who we are In *The Corrosion of Character*, Richard Sennett examined how flexible capitalism changes the experience of time, career and character. His contrast between a working life organized by durable commitments and another marked by projects, mobility and uncertainty shows that work provides more than income: it helps create biographical continuity. [W. W. Norton — The Corrosion of Character](https://wwnorton.co.uk/books/9780393319873-the-corrosion-of-character) AI intensifies the problem because it can shorten the half-life of skills, roles and routines. For a long period, a profession offered a relatively legible narrative: learn, enter, practice, accumulate experience, assume responsibility, transmit knowledge. That story was never universal or stable for everyone, but it became a central reference point of industrial society. When tools change quickly and parts of work can be automated, a career can become a sequence of adaptations. Optimistic language calls this “continuous learning.” The phrase can hide an asymmetry: a person is required to rebuild their market value repeatedly while firms own assets, data, platforms and capital that accumulate value over time. If the individual must be permanently flexible while infrastructure can remain permanently proprietary, flexibility stops being only freedom. It can become a transfer of risk. The social question is not merely whether someone “can adapt.” It is how much stability a society considers legitimate so that people can plan family, community, housing, education and a future. ## Keynes: if productivity frees us from work, will we know what to do with freedom? In 1930, John Maynard Keynes wrote *Economic Possibilities for our Grandchildren*. He imagined that technical progress could raise living standards enough to sharply reduce the amount of time required to satisfy economic needs. The essay became famous for imagining a much shorter workweek and for asking how human beings would live when necessity occupied less of life. [The Economics Network — Economic Possibilities for our Grandchildren](https://www.economicsnetwork.ac.uk/archive/keynes_persuasion/Economic_Possibilities_for_our_Grandchildren.htm) Economic output rose enormously after Keynes wrote. Working time fell much less than the technological possibility he imagined. Later scholarship on the essay notes that Keynes underestimated problems of distribution, inequality and persistent consumer desire. [MIT Press / Oxford Academic — Revisiting Keynes](https://academic.oup.com/mit-press-scholarship-online/book/29419) AI reopens Keynes’s problem in a new form. If a task that required eight hours can be done in two, there are at least four institutional paths: - produce the same output in less time and return hours to people; - produce four times more with the same number of people; - maintain output and reduce the number of workers; - invent new tasks, products and desires that absorb the released time. Technology does not choose among these paths. Modern history shows a strong ability to convert productivity gains into higher levels of production and consumption. Technical abundance therefore does not automatically create an abundance of time. Perhaps the most persistent scarcity is not material. It may be the institutional inability to say “enough.” ## Social exposure is unequal before any complete transformation occurs Philosophical reflection matters because effects do not arrive in an abstract society. They move through different classes, genders, countries and occupations. The International Labour Organization estimates that roughly one quarter of workers worldwide are in occupations with some degree of generative-AI exposure, while the share in the highest exposure gradient is much smaller. High exposure is greater among women, especially in high-income economies, because of occupational composition and the weight of clerical and administrative work. [ILO — Generative AI and Jobs](https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure) ```chart type: bar title: Employment share in the highest generative-AI exposure gradient unit: % Women · world | 4.7 Men · world | 2.4 Women · high-income countries | 9.6 Men · high-income countries | 3.5 ``` The geographic categories overlap and should not be added together. Exposure also does not mean replacement; it measures compatibility between occupational tasks and current AI capabilities. The chart matters less as an unemployment forecast than as a reminder of a historical rule: technologies arrive inside preexisting social structures. If an occupation is feminized, precarious, licensed, unionized, concentrated inside large firms or distributed among small businesses, the same technical capability will have different consequences. There is no “impact of AI” separate from institutions. ## Nine books, nine questions that remain open | Author and work | Central question | What it illuminates in the AI era | | --- | --- | --- | | Hannah Arendt — *The Human Condition* | what distinguishes necessity, world-building and action among people? | not every valuable activity is an automatable task | | Karl Polanyi — *The Great Transformation* | what happens when markets reorganize social relations? | economic efficiency can produce forms of insecurity that prices do not measure | | Karl Marx — *Grundrisse* | what happens when social knowledge becomes embodied in fixed capital? | cognitive productivity can be collective while ownership remains concentrated | | Norbert Wiener — *The Human Use of Human Beings* | who defines objectives, feedback and control in automatic systems? | automation is an institutional architecture, not only substitution | | Jacques Ellul — *The Technological Society* | what happens when efficiency becomes a dominant principle? | optimization can colonize spheres governed by other values | | Lewis Mumford — *The Myth of the Machine* | how do technique, organization and power form a social machine? | AI depends on an institutional infrastructure much larger than the model | | Ivan Illich — *Tools for Conviviality* | does the tool expand autonomy or dependence? | the relevant criterion is human capability after use | | Richard Sennett — *The Corrosion of Character* | how does economic instability affect biographical continuity? | permanent adaptation can transfer risk to individuals | | John Maynard Keynes — *Economic Possibilities for our Grandchildren* | what will we do when productivity reduces the need for work? | technical abundance requires institutions able to distribute time and income | The questions do not converge on one ideology. They converge on one principle: technologies should be judged partly by the social world they help build. ## The mistake of imagining intelligence and judgment are the same thing A society fascinated by AI can reduce intelligence to what benchmarks can demonstrate: answering, classifying, predicting, programming and synthesizing. Those capabilities matter economically. Judgment includes something else: deciding which problem deserves attention, accepting responsibility, understanding unmeasured consequences and knowing when a rule should not be applied. Judgment is not only an individual cognitive ability. It is a social position. A judge, physician, responsible engineer, teacher or public administrator is not legitimate merely because they produce good answers. They occupy institutions defining duties, procedures, appeal and accountability. If an AI produces a technically superior decision, a political question follows: who can challenge it? Who explains it? Who repairs harm? Who has authority to make an exception? Automation can reduce the cost of decisions while increasing the value of responsibility. ## Work can become less necessary before it becomes less morally central Modernity made employment both an income mechanism and a source of moral legitimacy. Everyday questions such as “what do you do?” often mean “what is your occupation?” Income without work is treated differently depending on its source: dividends can signal status while public assistance can carry stigma. That morality creates tension when machines perform more production. If less human work is economically required, a society may still insist that everyone prove usefulness through employment. The result can be functions with little meaning, artificially intensified targets or competition for positions that distribute income and status more than they satisfy productive necessity. An alternative requires separating three problems now bundled into employment: 1. **production:** how goods and services are created; 2. **distribution:** how income and property are allocated; 3. **recognition:** how people receive status, purpose and belonging. As long as employment solves all three, automation will look like liberation and threat at the same time. ## If code becomes cheap, society may discover that trust is expensive Abundant content and software move scarcity elsewhere. When producing text is costly, text has value partly because production is difficult. When images are costly, technical ability filters producers. When programming requires many hours, software embodies labor scarcity. AI reduces some of these costs. It does not eliminate value; it changes where value sits. The easier production becomes, the more important the following may become: - verifiable origin; - reputation; - curation; - accountability; - relationships; - distribution; - reliable data; - security; - physical presence; - community belonging; - the capacity to say “no.” ```mindmap What may remain scarce when applied intelligence becomes abundant - Trust - reputation - accountability - shared history - Attention - human time - focus - silence - Presence - care - relationship - physical experience - Power - ownership - infrastructure - access - Judgment - context - prudence - legitimacy - Meaning - purpose - identity - belonging ``` The AI economy may be less an economy without scarcity than an economy in which scarcity changes location. ## Education faces a contradiction efficiency cannot solve If a machine can write a better essay than a student, what is the purpose of assigning an essay? The answer depends on distinguishing product from formation. When a child solves an equation, writes a text or learns to program, the value is not only in the output. The activity changes the person. Difficulty, memory, repetition, error, frustration and revision are part of formation. A technology that delivers the product can destroy the process if used without judgment. This does not imply banning AI. Illich offers a better lens: the tool should expand the student’s ability to think rather than merely replacing the need to think. Education may need to become less centered on artifacts that can be generated cheaply and more centered on explanation, oral defense, investigation, experimentation, construction, interpretation, cooperation and responsibility. Paradoxically, AI may make teachers more important precisely because information becomes less scarce. The scarce resource becomes guidance about what is worth learning, how to distinguish understanding from fluency and how to transform information into judgment. ## The politics of AI begins where the efficiency question ends A firm has an economic reason to automate a task when doing so reduces cost, raises quality or creates new capacity. A society cannot use exactly the same criterion for every decision. Some activities derive part of their value from being human. Elder care, early-childhood education, justice, ritual, political representation, friendship, artistic creation and the response to suffering are not merely channels for producing outcomes. They are relationships. Automation can assist each one. But a society that substitutes relationships with efficiency can save resources while losing the content of the activity itself. The boundary between assistance and substitution will be political because it contains competing ideas of the good life. ## The deepest monopoly may be the power to define operational reality Discussions of concentration usually focus on market share. There is another form of power: setting the standards through which organizations perceive and act. If a small number of platforms provide models, identity, cloud infrastructure, memory, payments, enterprise tools and distribution channels, they can shape the default architecture of economic activity without directly controlling every firm. Ellul would focus on standardization through efficiency. Mumford would see the formation of an institutional megamachine. Polanyi would ask how social relations are redesigned through markets. Marx would ask who owns objectified knowledge. Illich would ask whether users can exit without losing capability. Relevant concentration is not only “how many companies exist?” It is **how much of economic life must pass through an architecture defined by a small number of institutions**. ## Freedom is not having more options; it is retaining the capacity to refuse Technology usually increases options: more information, more products, more speed, more access. Freedom also has another dimension: being able to reject an option without being excluded from social life. If every job requires one system, every school one platform, every business one cloud and every conversation intermediaries, many features can coexist with little structural autonomy. Illich helps formulate a simple test: a good tool should allow use without requiring total submission to its logic. That implies preserving interoperability, human knowledge, local alternatives, data portability, rights to challenge automated decisions and a real capacity to operate when central systems fail. Redundancy, viewed by efficiency as waste, can be a condition of freedom. ## A cognitively abundant society can still be deeply unequal The word “abundance” can mislead. If cheap intelligence is available to everyone, small firms, students and workers can acquire capacities once reserved for large organizations. That is democratizing. But if the infrastructure producing that intelligence is highly concentrated, part of the gains can flow to owners of computing, electricity, data and distribution. The same technology can reduce inequality of capability while increasing inequality of wealth. There is no contradiction. Smartphones made previously expensive tools available to billions while concentrating enormous value in a small number of platforms. The political problem becomes how to distribute not only access to the service but participation in productivity gains. ## The future of work may be a dispute over who receives the time that productivity releases Productivity gains create potential time. A firm can capture it as higher output. A consumer can capture it as a lower price. A shareholder can capture it as margin. A worker can capture it as higher pay or shorter hours. The state can capture part of it as tax revenue and public service. The distribution of this invisible time may become one of the central social conflicts of automation. Keynes imagined productivity freeing people from economic necessity. Sennett reminds us that work also organizes biography. Arendt reminds us that release from labor does not automatically produce meaningful public action. A society needs to know what to do with time if less work is not to mean less belonging. ## Four social responses to abundant intelligence These are not forecasts. They are possible institutional forms. | Response | Organization of work | Distribution | Likely human effect | | --- | --- | --- | --- | | intensification | AI raises targets and pace without reducing hours | most gains accrue to firms and owners | higher productivity with pressure and insecurity | | substitution | automation reduces the need for work and employment | income remains heavily tied to employment | productive abundance with distributional conflict | | democratization | cheap tools expand the capacity of individuals and small organizations | competition and ownership become more distributed | more autonomy, creation and mobility | | liberation of time | productivity is partly converted into shorter hours and universal services | institutions distribute part of the social gain | work becomes less central without necessarily losing meaning | Different sectors can combine all four paths. ## What does it mean to remain human when producing answers stops being special? For centuries, much intellectual prestige was attached to the ability to produce scarce answers: calculate, translate, memorize, write, diagnose, program. When machines perform some of these activities, a pessimistic interpretation concludes that human beings have “lost value.” That conclusion only follows if human value was equivalent to the scarcity of a productive skill. Arendt points elsewhere: humanity appears in the capacity to begin something among others. Illich emphasizes autonomy. Wiener asks about purpose. Sennett reminds us of the need to build a life story. Polanyi shows that society cannot be reduced to the market. Even Marx, focused on production, saw that the development of productive forces changes the conditions of emancipation. AI may not diminish the human. It may expose how much we had confused human value with economic utility. ## The final question is not whether machines will become human The more important question is whether human institutions will become machines. A firm can treat every decision as optimization. A school can treat all learning as measurement. A government can treat a population as a set of variables. A person can treat their own life as a productivity dashboard. None of this requires artificial consciousness. A society can become culturally mechanized before a machine becomes person-like. That is where the philosophical challenge of AI lies. The more powerful the tools, the less adequate it is to evaluate the future only by what they can do. The criterion must also include what they do to us, what we begin to expect from one another and which forms of life disappear when everything that is not efficient starts to look irrational. ## Conclusion: technical abundance does not solve the human question Artificial intelligence can extend one of history’s greatest achievements: making knowledge, production and creative capacity accessible to more people with less effort. It can also deepen an older tendency: organizing more dimensions of life according to efficiency criteria defined by systems controlled by relatively few institutions. There is no contradiction between those possibilities. They can occur at the same time. Arendt shows that living is more than laboring. Polanyi shows that markets need to remain embedded in society. Marx shows that collective knowledge can become concentrated productive power. Wiener shows that automation is a problem of purpose and control. Ellul shows that efficiency can stop being a means and become a system. Mumford shows that machines are also organizations. Illich shows that tools should be judged by the autonomy they leave behind. Sennett shows that work structures time and character. Keynes shows that productivity becomes freedom only if a society can transform abundance into human time. The AI era brings these questions into ordinary life. It is not enough to ask how many jobs will disappear. We must ask how people will receive income when less work is required. It is not enough to ask how much software will be produced. We must ask who controls the systems through which that software circulates. It is not enough to ask whether students will have better answers. We must ask whether they will continue learning how to formulate questions. It is not enough to ask whether decisions become more efficient. We must ask who has authority to challenge them. It is not enough to ask whether intelligence becomes abundant. We must ask **whether freedom, autonomy, trust, time and belonging become abundant as well**. Technology can reduce the scarcity of answers. The social task remains deciding which questions deserve to organize a civilization. ## Books and reference sources - Hannah Arendt, *The Human Condition*. [University of Chicago Press](https://press.uchicago.edu/ucp/books/book/chicago/H/bo29137972.html) - Karl Polanyi, *The Great Transformation: The Political and Economic Origins of Our Time*. [Beacon Press](https://www.beacon.org/The-Great-Transformation-P46.aspx) - Karl Marx, *Grundrisse: Foundations of the Critique of Political Economy*. [Penguin Classics](https://www.penguin.co.uk/books/35199/grundrisse-by-karl-marx-translated-with-a-foreword-by-martin-nicolaus/9780140445756) - Norbert Wiener, *The Human Use of Human Beings: Cybernetics and Society*. [Google Books bibliographic record](https://books.google.com/books/about/The_Human_Use_Of_Human_Beings.html?id=DydKDgAAQBAJ) - Jacques Ellul, *The Technological Society*. [Google Books](https://books.google.com/books/about/The_Technological_Society.html?id=TEs0AAAAMAAJ) - Lewis Mumford, *The Myth of the Machine*. [Google Books](https://books.google.com/books/about/The_Myth_of_the_Machine.html?id=FtIyAAAAMAAJ) - Ivan Illich, *Tools for Conviviality*. [Google Books](https://books.google.com/books/about/Tools_for_Conviviality.html?id=EgKaPwAACAAJ) - Richard Sennett, *The Corrosion of Character: The Personal Consequences of Work in the New Capitalism*. [W. W. Norton](https://wwnorton.co.uk/books/9780393319873-the-corrosion-of-character) - John Maynard Keynes, “Economic Possibilities for our Grandchildren,” in *Essays in Persuasion*. [The Economics Network](https://www.economicsnetwork.ac.uk/archive/keynes_persuasion/Economic_Possibilities_for_our_Grandchildren.htm) - Stanford Encyclopedia of Philosophy, “Philosophy of Technology”. [Stanford Encyclopedia of Philosophy](https://plato.stanford.edu/entries/technology/) - International Labour Organization, *Generative AI and Jobs: A Refined Global Index of Occupational Exposure*. [ILO](https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure)