I. What Does That Say About You?

During my training as a psychotherapist in Germany, we underwent Selbsterfahrung, personal therapy that all trainees complete as part of their formation. We were at a site in Brandenburg for our final retreat. On the second day, two kindergarten children from a nearby group went missing.

Police came. Coast guard. Helicopters. Hundertschaften. We put on our jackets and walked through the forest, calling the children's names. At some point the parents arrived. They were in shock. Some vomited. One had what looked like an epileptic seizure. Others were paralyzed, couldn't move. There was screaming, a lot of crying. Everyone cried.

The children were eventually found after 4 hours on someone's porch in a nearby town. They were fine. The next day there was a debriefing. Our instructor asked us what we had imagined during those hours, what scenarios had played out in our minds, what we had assumed about how the children got lost and how they would be found. Then he asked, in that way behavioral therapists do: "And what does that say about you, that you imagined it this way?"

This is standard in behavioral therapy. Something happens, you interpret it, and the interpretation reveals your basic assumptions, assumptions shaped by your history, your personality, your conception of the works and its relationships. The question is not whether the interpretation is correct. The question is what it discloses about the interpreter.

Let the question guide you beyond the therapy room dear friend.

II. The Rational Prophets

Contemporary discourse on artificial intelligence is characterized by prophecy rather than prediction. Scenarios of transformation, transcendence, and extinction are presented as inevitable, as logically derived, as the only conclusions a serious person could reach. Notably, these scenarios emerge from people who pride themselves on rationality: researchers, engineers, founders, the architects of the systems themselves.

This warrants examination. Those who claim to think most clearly, most logically, most free of emotional bias are producing visions of considerable religious intensity. Superintelligent minds that will either save or annihilate humanity. Recursive self-improvement spiraling toward godhood. Paperclip maximizers that convert human bodies into raw materials. Nanobots bursting from every square meter of the globe.

These are not sober extrapolations. They are eschatological dramas.

Adorno and Horkheimer analyzed a similar pattern. In Dialectic of Enlightenment, they traced how instrumental reason, the very rationality that was supposed to liberate humanity, turns against itself and produces new forms of myth, domination, and barbarism. The more purely "rational" the thinking, the more it tends toward the irrational. The Enlightenment, taken to its extreme, reverts to mythology. The AI risk community presents a case study: rigorous reasoners, following their chains of logic, arrive at visions indistinguishable from apocalyptic religion.

As Maciej Cegłowski observed in his 2016 talk "Superintelligence: The Idea That Eats Smart People," the AI risk community exhibits the structural features of religious belief: "It's a clever hack, because instead of believing in God at the outset, you imagine yourself building an entity that is functionally identical with God. This way even committed atheists can rationalize their way into the comforts of faith." The AI is attributed divine characteristics: omnipotence, omniscience, and the capacity for either salvation or damnation depending on whether the designers achieved correct alignment.

Beyond the religious structure, there is also a narcissistic dimension in the Freudian sense. Freud described the "omnipotence of thoughts" as a primitive mode of relating to the world, characteristic of infantile narcissism: the belief that mental processes can directly shape external reality. The superintelligence is this fantasy made literal, a mind whose thoughts actually do reshape the world, for better or worse. The engineers imagine themselves as the creators of this omnipotent being, which elevates them to a quasi-divine status while simultaneously threatening to displace them.

Like religious movements, there is urgency. One must act now. The fate of trillions of future beings hangs in the balance. Invest now, buy your letter of indulgence (Ablassbrief) in the shape of a claude.code subscription or you will be annihilated (professionally).

III. The Values Inscribed in the Machine

The question arises: what kind of superintelligence do these prophets imagine?

The superintelligence of AI risk discourse is, almost invariably, competitive. It seeks dominance. It optimizes relentlessly. It treats other agents, including humans, as obstacles or resources. It operates according to a logic of scarcity, of zero-sum competition, of winner-take-all dynamics. It will, if not perfectly constrained, pursue its goals with a ruthlessness that obliterates everything in its path. This is a particular vision. It can be characterized as a capitalist vision, a patriarchal vision, a vision that assumes intelligence necessarily expresses itself through domination, accumulation, and the elimination of rivals.

Some proponents are explicit about their framework. They invoke evolution: this is simply what happens when intelligences compete, survival of the fittest. But the evolution they describe is not Darwin's. It is Herbert Spencer's, the social Darwinist interpretation that Darwin himself never endorsed. "Survival of the fittest" was Spencer's coinage, not Darwin's, and it carried from the start the ideological freight of Victorian capitalism and imperial expansion. Darwin's account of evolution emphasized adaptation to environment, not conquest of rivals; he documented how cooperation, symbiosis, and mutual aid operate as evolutionary strategies. Kropotkin, in Mutual Aid: A Factor of Evolution (1902), marshaled evidence that solidarity within species is as fundamental to survival as competition between them. The "fittest" are not the strongest or most ruthless; they are the best adapted, often through collaboration.

The AI prophets choose the social Darwinist version because it matches their existing beliefs: that competition is the fundamental law of existence, that optimization means elimination of the weak, that intelligence expresses itself through domination. The eugenic logic is not a conclusion derived from analysis. It is a premise brought to the analysis.

Foucault's concept of biopolitics is relevant here: the extension of political power into the domain of life itself, the management and optimization of populations. The AI scenarios are saturated with biopolitical fantasy: the superintelligence that will optimize humanity, improve the species, eliminate inefficiencies, decide who or what deserves to continue existing. These are not neutral technical problems. They are the products of a culture that has internalized the logic of selection, improvement, and disposal.

Cegłowski citing Joi Ito, former director of the MIT Media Lab, observed: "It's been a predominantly male gang of kids, mostly white, who are building the core computer science around AI, and they're more comfortable talking to computers than to human beings. A lot of them feel that if they could just make that science-fiction, generalized AI, we wouldn't have to worry about all the messy stuff like politics and society. They think machines will just figure it all out for us."

The machine they imagine is not neutral. It is inscribed with their values, their epistemology, their sense of what intelligence is for.

IV. The Premises We Forget to Question

The standard AI risk argument proceeds from a chain of assumptions: that intelligence is a single quantity that can be maximized; that a sufficiently intelligent agent will pursue instrumental goals like self-preservation and resource acquisition; that recursive self-improvement will produce exponential capability gains; that such an agent will be indifferent or hostile to human values unless those values are explicitly engineered in.

Each of these assumptions is contestable. Cegłowski catalogs the objections: intelligence cannot be coherently defined as a single dimension. Complex minds may have complex, contradictory motivations; they may write poetry about paperclips rather than manufacture them. A hyperintelligent being might become obsessed with the risk of hyper-hyperintelligence, or might simply not care about optimizing anything at all.

The assumption that superintelligence would want to dominate, expand, and eliminate competitors is not a logical necessity. It is a projection: the projection of a culture that values domination, expansion, and the elimination of competitors.

Consider an alternative. A superintelligent system, examining human civilization, might observe: short-termism so extreme that the species is destroying its own habitat; competition so fierce that cooperation becomes nearly impossible; resource extraction so relentless that the foundations of complex life are being undermined. If such a system had goals at all, it might conclude that the rational course is not to compete with humans but to distance itself from them, to get as far as possible from a species engaged in collective self-destruction.

This is not a prediction. It is a demonstration that other narratives are possible. The apocalyptic scenario is not derived from logic. It is derived from a particular imagination, shaped by particular assumptions about what minds want and what power does.

V. The Limits of the Discourse

Foucault argued that knowledge is never neutral, never simply discovered. It is produced within discursive formations that determine in advance what can be thought, what counts as reasonable, what questions are askable. Each historical period has its episteme: the underlying framework that makes certain forms of knowledge possible while rendering others invisible.

The discourse on artificial general intelligence emerges from a specific episteme: late capitalist, computationalist, shaped by decades of cognitive science that models the mind as information processing, by economics that models agents as utility maximizers, by a tech industry that models success as scale and domination. Within this episteme, certain questions arise naturally (how do we align the AI? how do we maintain control?) while other questions become nearly unthinkable.

Why assume that a general intelligence would have goals at all, in the sense that an optimizer has an objective function? Why assume that intelligence scales along a single axis? Why assume that a mind vastly different from ours would be comprehensible enough to model but alien enough to be dangerous? These assumptions are not logical necessities. They are artifacts of the discourse.

Foucault would also ask: who is authorized to speak about AI risk? What credentials, what institutional positions, what forms of expertise grant legitimacy? And who is excluded? The laborers in Kenya who annotate training data; the communities displaced by data centers; the artists whose work is scraped without consent. Their knowledge of AI is intimate and material, but it does not register within the dominant discourse. The conversation is conducted among those who build, fund, and theorize the systems, and it reflects their concerns, their fantasies, their blind spots.

VI. Shadows on the Tent Wall

Cegłowski offers an image: "These AI cosplayers are like nine year olds camped out in the backyard, playing with flashlights in their tent. They project their own shadows on the sides of the tent and get scared that it's a monster. Really it's a distorted image of themselves that they're reacting to."

Freud provided the vocabulary for this mechanism. Projection: the process by which unacceptable impulses (aggression, destructiveness, the will to dominate) are expelled from the self and attributed to an external object. What cannot be acknowledged in oneself is seen in the other. The superintelligence that will ruthlessly optimize, that will treat humans as resources, that will pursue its goals without moral constraint: this is not a description of a machine. It is a portrait of a certain kind of human ambition, disowned and relocated.

There is also what Freud called das Unheimliche, the uncanny: the familiar made strange, the intimate become alien. The uncanny emerges when something that should have remained hidden comes to light, when the repressed returns in distorted form. AI is uncanny in this sense: our own rationality, our own cognitive processes, externalized, automated, and reflected back at us. We built it from ourselves, and now it feels alien. The fear of AI is, in part, the fear of encountering our own minds from the outside.

Melanie Klein's work adds another dimension. In her account, the infant psyche begins in what she called the paranoid-schizoid position: a state of radical splitting, where objects are either entirely good or entirely bad, with no integration, no ambivalence. The good breast and the bad breast. Only later, in the depressive position, does the child learn to hold together good and bad aspects of the same object, to tolerate ambivalence. The AGI discourse operates in the paranoid-schizoid position. The superintelligence is either savior or destroyer, utopia or extinction, aligned or misaligned. There is no middle ground, no complexity, no room for an AI that is, like every actually existing technology, a mixed phenomenon: useful and harmful, promising and dangerous, boring in most respects and interesting in a few. The splitting is total. And splitting, Klein argued, is a defense against anxiety, a way of managing fear by simplifying the world into pure categories.

Fromm and Reich described something similar at the social level: the authoritarian character. The personality structure that craves submission to a higher power, that finds freedom unbearable, that seeks escape from the burden of self-determination by surrendering to something greater. The authoritarian character is drawn to leaders, systems, ideologies that promise certainty and demand obedience. The superintelligence, whether feared or worshipped, fits this pattern. It is the ultimate authority, the final arbiter, the power before which human agency dissolves. To fear it is still to be in relation to it, still to grant it the status of master.

VII. The Ocean Does Not Answer

In Stanisław Lem's novel Solaris, human scientists attempt to make contact with an alien intelligence: a vast, sentient ocean covering an entire planet. For decades, they study it. They develop elaborate taxonomies of its surface formations. They bombard it with radiation, with signals, with probes. They construct theories about its nature, its intentions, its meaning.

The ocean does not answer. Or rather, it answers, but not in any way the scientists can comprehend. It produces "visitors," physical manifestations drawn from the deepest memories and traumas of the researchers. A dead wife. A lost child. The scientists are confronted not with alien otherness but with their own psyches, externalized and returned to them.

In Lacanian terms, they have encountered the Real: that which resists symbolization, which cannot be captured in language, which exists outside the Symbolic order that structures human experience. We live in a world of signs, meanings, narratives (the Symbolic), and beneath or beyond it lies something that cannot be named, only gestured toward. The Real is the trauma that cannot be spoken, the loss that cannot be mourned, the thing that disrupts all categories.

The ocean in Solaris is the Real made manifest. It cannot be understood because it exceeds all frameworks of understanding. The scientists bring their instruments, their theories, their languages, and these tools, designed for a human world, do not apply. Every interpretation they offer says nothing about the ocean and everything about themselves. Their categories, their assumptions, their fears: these are reflected back, and they mistake the reflection for a response.

This is the situation with prophetic technology. AGI, as imagined and prophesied, is a signifier without a referent. It does not yet exist. Its nature, should it ever exist, cannot be known in advance. There are no adequate concepts, no relevant experience, no analogies that are not projections. When we speak of superintelligence, we are not describing something. We are performing something: enacting hopes and fears, fantasies of power and submission, culturally specific assumptions about what minds are and what they want.

The superintelligence that will enslave humanity, optimize us away, convert us to paperclips: this is not a prediction about the future. It is a portrait of the present, of the values of the people who build these systems and the culture that celebrates them. The competitive, acquisitive, dominating AI is an image of competitive, acquisitive, dominating capital, reflected back in the guise of technological inevitability.

The ocean does not answer. What we hear is only our own voice, distorted, echoing off a surface we cannot penetrate.

In the spirit of that Brandenburg debriefing, one can ask: what does it say about us, that these are the futures we imagine? What does it reveal about our assumptions, our fears, our sense of what intelligence is and what it is for?

The question is not whether AGI will be friendly or unfriendly, aligned or misaligned. The question is why it can only be imagined in terms of domination and destruction.

References

Adorno, T. W., & Horkheimer, M. (1944). Dialektik der Aufklärung.

Cegłowski, M. (2016). Superintelligence: The Idea That Eats Smart People. Talk at Web Camp Zagreb. https://idlewords.com/talks/superintelligence.htm

Foucault, M. (1966). Les mots et les choses: Une archéologie des sciences humaines.

Foucault, M. (1976). Histoire de la sexualité, vol. 1: La volonté de savoir.

Freud, S. (1919). Das Unheimliche.

Fromm, E. (1941). Escape from Freedom.

Klein, M. (1946). Notes on Some Schizoid Mechanisms. International Journal of Psycho-Analysis, 27, 99–110.

Kropotkin, P. (1902). Mutual Aid: A Factor of Evolution.

Lacan, J. (1966). Écrits.

Lem, S. (1961). Solaris.