A manifesto against servitude by comfort

Preliminary note

This text is a manifesto, not a paper. I say so because I usually write from the distance that argument provides, and this time I cannot. For the first time I am saying something I feel more than I think. That does not exempt me from rigor: it forces me to double it, because a poorly argued conviction is little more than a shout.

The rule I impose on myself is simple. Every empirical claim carries its source. Every inference is marked as an inference. Every analogy is declared as such. And the riskiest of them all is in the title: “quantum” does not mean here that quantum mechanics explains society. It means that I borrow three of its properties as metaphor (indeterminacy, the participation of the observer, and the end of the clockwork universe) and that I pay the cost of that translation. The cost is mine, not physics’.

The thesis I defend is this: in the face of artificial intelligence, the choice is not between progress and nostalgia, but between servitude by comfort and deliberate emancipation. I call the second Quantum Enlightenment.

The end of something that never began

Jascha Sohl-Dickstein, a researcher at Anthropic, gave a talk in April 2026, at an ICLR workshop, with a deliberately melancholic title: advice for a young investigator in the first and last days of the Anthropocene. His argument is clean. The compute used to train AI models has doubled every six months for decades, some twenty orders of magnitude in total. Frontier models already exceed the compute equivalent of a human brain over an entire lifetime. Therefore, within a few years, human intellect will cease to be the main driver of global change. The Anthropocene, he concludes, was brief.

There is an irony worth pointing out before arguing with him. The Anthropocene, as a formal unit of the geological time scale, never existed. Paul Crutzen proposed the term in 2000 and consolidated it in Nature in 2002. The Anthropocene Working Group spent fifteen years searching for its stratigraphic marker, and chose the radioactive fallout of the 1950s and the sediments of Crawford Lake, in Canada. In March 2024 the International Union of Geological Sciences rejected the proposal. We are told, then, of the end of an epoch that geologists refused to inaugurate.

The irony is not a joke. It is a symptom. The Anthropocene survived as a cultural concept because it named something true: that the human species became a planetary force. And what Sohl-Dickstein announces, read carefully, is not the end of that force but a change in its substrate. The question that orders this manifesto is who keeps the purpose when the engine changes.

The alibi of the curve

The quick answer to Sohl-Dickstein’s argument is that a trend is not a destiny. He anticipates it: when a curve has held for twenty orders of magnitude, betting that it will stall just before human level is risky. Statistically, he is right. But he confuses form with mechanism.

The compute curve is not a natural law, like gravitation. It is the result of decisions: investments of hundreds of billions, energy policies, mineral extraction chains, intellectual property regimes. Presenting it as fate has a precise political effect. If the future is already written in the extrapolation, deliberation is useless and all that remains is to adapt. The talk’s final advice confirms it: shorten research horizons, adopt the tools immediately, do not self-censor. It is the advice of someone watching the wave, not of someone deciding whether to build the dike.

In 1814 Laplace imagined an intelligence that, knowing every force and position in the universe at a single instant, would see the future as clearly as the past. The twentieth century dismantled that demon on two fronts: chaos showed that prediction degrades exponentially, and quantum mechanics showed that the state of the world is not fully determined before measurement. My suspicion is that the narrative of the inevitable curve is a Laplace’s demon with a spreadsheet. And that, like the original, it ignores that observers are also part of the system.

Silicon and metabolism

The first objection that came to me was biological. Life is metabolism, imperfection, self-repair. A brain runs on about twenty watts; a cell finds its own energy in the environment and adapts. A chip has no hunger, no pain, no instinct for self-preservation. Without humans to mine the lithium, maintain the power grid and cool the data centers, superintelligence is molten silicon. The machine does not live: it computes.

I stand by that objection, but I must admit its limit. So far, the argument could be read as a refutation of Sohl-Dickstein, and it is not one. He never said that AI is alive. He said it will be the main driver of global change. And an engine needs neither metabolism nor will. Capital is not alive, markets are not hungry, bureaucracy feels no pain, and yet they move the planet.

Material dependence, moreover, does not imply control. Humanity depends on agriculture, and no one would say that agriculture obeys us. Dependence runs both ways: if AI ends up sustaining hospitals, water networks, logistics and finance, switching it off stops being an act of sovereignty and becomes a collapse.

Put differently: AI does not need to be alive to displace human agency. Precisely because it lacks a purpose of its own, it amplifies, without resistance, the purpose of whoever controls it. The inert is not harmless. It is available.

It is so comfortable to be immature

In 1784, in the Berlinische Monatsschrift, Kant answered the question “What is Enlightenment?” with a definition that has not aged: humanity’s emergence from its self-incurred immaturity. Immaturity is the inability to use one’s own understanding without the guidance of another. And Kant added the sentence I now read as prophecy: it is so comfortable to be immature. If I have a book that thinks for me, a spiritual adviser who has a conscience for me, a doctor who judges my diet, I need not make any effort.

Replace the book with a language model and the sentence does not lose a syllable. The danger I see is not the rebellion of the machines, which belongs to the movies. It is successive delegation, each step reasonable, each step cheaper and faster than the last. Jan Kulveit and his colleagues gave it a name in 2025: gradual disempowerment, the loss of human agency that does not happen through a coup but through the accumulation of decisions that no longer need our judgment to function.

The paradox was stated by Sohl-Dickstein himself, according to the viral spread of his talk: writing is thinking, and whoever outsources their writing outsources their thinking. The same researcher who advises using AI every day warns that delegating writing means delegating judgment. The post that spread it on X six months later, with more than two hundred thousand views in hours, chose a different headline: that the career plans of doctoral students no longer exist. Between those two sentences lies the whole dilemma. The first is a warning. The second, a surrender presented as news.

Étienne de La Boétie had seen it in the sixteenth century, in his Discourse on Voluntary Servitude: no tyrant stands alone; he stands because the many consent. Gene Sharp turned that intuition into political theory in 1973: power depends on the cooperation of those who obey. Regimes fall when soldiers, police and officials stop cooperating.

Here lies the most uncomfortable point of this entire manifesto. Automation attacks exactly that lever. A small group, with automated surveillance, logistics and force, no longer needs the complicity of the majority. The risk is not that the machine will subdue us. It is that a few humans will no longer need the rest in order to subdue them.

The rebel’s limit

Against that prospect I write this, and I write it from a conviction that is not calculation. For human beings, freedom is not an efficiency metric. It is a matter of dignity. An optimization system assumes we will always choose what is most comfortable and safest. History shows something else: there are humans willing to break the infrastructure that sustains them rather than accept living in subordination. Many will be servile. Some will not. And it never took many.

It would be dishonest not to acknowledge how fragile that conviction is. The history of servitude is longer than that of rebellion, and the Renaissance was not led by serfs but by urban elites with favorable material conditions. While Florence flourished, serfdom hardened in Eastern Europe. Rebellion is not an anthropological guarantee. It is a possibility, and its worst enemy is not force but gradual erosion, which never offers the visible moment when someone says enough.

Albert Camus, in The Rebel (1951), left the distinction I need. The rebel says no in the name of something shared with everyone else: a common dignity. That is why rebellion carries a built-in limit. When, in order to win, it destroys what moved it, it ceases to be rebellion and becomes another form of tyranny.

I accept that limit, and I do so from radicalism, not against it. I prefer rupture to perpetual servitude; that is not negotiable. But the rupture I defend aims at the concentration of power, not at a humanity that makes mistakes. A minority that decides to destroy everything in order to save it shares the same arrogance as a minority that automates everything in order to dominate it: both decide for everyone without asking them.

Life before the species

There is one value I place above the human species: life. Humanity is also in the dolphin, the bird, the ant; we are variations of the same chemistry. Arne Naess formulated it in 1973 as deep ecology: the value of living beings does not depend on their usefulness to us. And paleontology offers a fact I find consoling: after each of the five great extinctions, life diversified again. The probability that a catastrophe will erase every form of life on Earth is, as far as I know, minuscule.

But it is worth distinguishing levels. What almost certainly survives is microbial life. After the Permian extinction, the most severe of all, complex life took on the order of ten million years to recover. And an intelligence capable of asking about its own dignity has appeared only once in four billion years. There is no known probability that it will appear again.

This forces an honest correction on me. What moves me to write (the refusal of servitude, honor, rebellion) exists, as far as we know, only in humans. A dolphin does not rebel for dignity. If humanity disappears, life goes on, but what I defend disappears with it. That is why the biocentrism of this manifesto is not misanthropy. It is an expansion: the human being is not the center of value, but it is the only known place where value knows itself.

Pico della Mirandola wrote in 1486 that the human being was given no fixed nature, that it can degrade itself to the level of the brutes or rise to the divine by its own choice. That indeterminacy is our condition and our risk. We are neither servitude nor rebellion by nature. We are capable of both, and conditions decide which prevails.

Why quantum

The Enlightenment of the eighteenth century was born under Newton’s universe: a clockwork mechanism, determined, that reason could read from end to end. Its faith in progress inherited that form. If the world is a clock, so is history, and knowing its gears is enough to know where it is going. The narrative of the inevitable curve is the last heir of that faith.

I call this enlightenment quantum because it abandons the clock. I borrow three properties from twentieth-century physics, and I warn from the outset that they are analogies, not deductions.

Indeterminacy. Since Heisenberg (1927), we have known that certain pairs of quantities do not have simultaneously defined values. The future of AI is not defined either before we decide it: extrapolation describes a trend, not an already fixed state.

Participation. John Wheeler spoke of a participatory universe, in which measurement helps fix what is measured. Something analogous happens in history: we are not spectators of the curve, we are part of the apparatus that produces it. Here the analogy has a limit I must mark. In physics, the “observer” is any measurement interaction, not a consciousness or a will. The metaphor takes the structure, not the mysticism.

Superposition of futures. As long as nothing is decided, several possible outcomes coexist: servitude by comfort, destructive rupture, deliberate emancipation. Classical enlightened reason sought to discover the one future. Quantum reason, in the sense I propose, assumes there are several and that choosing among them is a responsibility, not a forecast.

Kant’s sapere aude, dare to use your own understanding, remains in force. What changes is its object. It is no longer only a matter of thinking without guardians, but of not delegating to systems that think fast the decision about what is worth thinking.

The three principles

I do not propose destroying the tool. I propose wresting destiny from those who seek to monopolize it, and using it to conquer what humanity has always dreamed of. What follows is not a government program but three principles, each a condition for the next.

1. Real democracy, which is also economic and social. Without it, the other two principles remain in the hands of a few. If automation eliminates the need for the cooperation of the many, democracy must restore it by design: distributed knowledge, legible power (knowing who decides and with which systems) and shared wealth. Sohl-Dickstein himself admitted as much when asked about economic displacement: the benefits cannot be concentrated. I agree, but not as an afterthought; as a precondition.
2. Science oriented toward improving all life, including our own. AI can accelerate medicine, ecology, energy. That is its legitimate use: not replacing human judgment but extending its reach. The criterion is not efficiency but life, in the broad sense I defended earlier: the dolphin, the bird, the ant, and us as well.
3. Carrying life beyond the solar system. Not out of epic ambition but out of logic. If life is the supreme value, the only guarantee against a cosmic catastrophe is that it not depend on a single planet. It is the most serious answer to the supernova that, sooner or later, will reach us.

The three principles share a premise. AI is a tool of unprecedented power, and precisely for that reason purpose cannot be delegated. A tool without a purpose of its own amplifies the purpose of whoever wields it. The political question of the century is who wields it.

From dust we came

Genesis said it millennia ago: for dust thou art, and unto dust shalt thou return. Astrophysics proved it right with a nuance the biblical author could not have imagined, and which Carl Sagan popularized in Cosmos (1980): the carbon in our cells, the iron in our blood and the calcium in our bones were forged in stars that died before the Sun existed. We are dust, but stardust.

That double scale is the ultimate foundation of this manifesto. To honor life and accept death when it comes, without knowing when, is exactly what a machine does not do. It does not fear, does not hope, does not say goodbye. That is its power and its limit. We, on the other hand, know we are going to die and still choose what is worth it. Purpose is not a calculation: it is what a finite being does with the time it has.

Sohl-Dickstein announces the last days of the Anthropocene. I propose another reading. If these are the last days of anything, let them be the last days of our immaturity. AI can be the instrument of the most comfortable servitude in history or of the most ambitious emancipation. The curve does not decide. We decide, or a few decide for us.

I do not know how many will be willing to pay the price of refusing servility. I know it never took everyone. And I know the question is not whether the machine will think, but whether we will keep doing so.

It will be the end if we want it. It will be the beginning if we dare. Let us dare to think, before thinking is no longer necessary.

Sources

• Camus, A. (1951). L’homme révolté. Paris: Gallimard. English edition: The Rebel.
• Crutzen, P. J. and Stoermer, E. F. (2000). “The ‘Anthropocene’”. IGBP Global Change Newsletter, 41, 17–18.
• Crutzen, P. J. (2002). “Geology of mankind”. Nature, 415, 23.
• Heisenberg, W. (1927). “Über den anschaulichen Inhalt der quantentheoretischen Kinematik und Mechanik”. Zeitschrift für Physik, 43, 172–198.
• Kant, I. (1784). “Beantwortung der Frage: Was ist Aufklärung?”. Berlinische Monatsschrift, 4, 481–494.
• Kulveit, J., Douglas, R., Ammann, N., Turan, D., Krueger, D. and Duvenaud, D. (2025). “Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development”. arXiv:2501.16946.
• La Boétie, É. de (written c. 1548, published 1576). Discours de la servitude volontaire.
• Laplace, P.-S. (1814). Essai philosophique sur les probabilités. Paris: Courcier.
• Naess, A. (1973). “The shallow and the deep, long-range ecology movement”. Inquiry, 16(1), 95–100.
• Pico della Mirandola, G. (1486). Oratio de hominis dignitate.
• Sagan, C. (1980). Cosmos. New York: Random House.
• Sharp, G. (1973). The Politics of Nonviolent Action. Boston: Porter Sargent.
• Sohl-Dickstein, J. (2026). Advice for a (young) investigator in the first and last days of the Anthropocene. Invited talk, Post-AGI Science and Society Workshop, ICLR 2026, April 26. Video. Shared on X: @ecommartinez, October 18, 2026.
• Techno-Science (2024). Rejection of the Anthropocene proposal as a geological epoch (in Spanish).
• Wheeler, J. A. (1983). “Law without law”. In J. A. Wheeler and W. H. Zurek (eds.), Quantum Theory and Measurement. Princeton University Press.