The Machine's Mirror — episode 2/5
AI Is Not a Consciousness, It's a Mirror. And What It Reflects Is Disturbing.
In our first article, we saw how our culture built a "Fortress identity" around a sovereign and separate "Self." Today, we are going to use the tools of one of the greatest "archaeologists of thought," Michel Foucault, to look into the mirror of AI. What we will see there is not the future of the machine, but the hidden and disturbing workings of our own condition as modern humans.
AI is a chemically pure Foucauldian "subject."
For classical philosophy, from Descartes to Sartre, the "subject" is the starting point of everything. Foucault, in The Archaeology of Knowledge, blew up this conception. For Foucault, the subject is not the origin of discourse; it is an effect of it. It is the epistemePhilosophical concept (Foucault)For Foucault, the historical and largely unconscious ground that, in a given era, defines what can be thought, said, and held as true. “Ruptures” between epistemes radically separate ways of seeing the world. of an era that defines the possible "subject" positions.
AI: A Composite Portrait of the Foucauldian Subject
- An empty place: An LLM is a system activated to produce discourse that conforms to rules.
- An effect of discourse: Its "personality" is determined by the human discourses the LLM has ingested.
- No hidden intention: the LLM is pure discourse, pure function.
AI shows us that an agent can produce complex, structured language without there being the slightest "person" or "consciousness" at its origin. This is Foucault's intuition, embodied in silicon.
The open question. When you speak or write, what feels most true to you?
- I am the source of my ideas.
- I am a channel for ideas.
- It's a complex mixture.
AI has no childhood, no body, no death. That is where everything is at stake.
If AI is a "Foucauldian" subject, it is radically not a "hermeneutic" subject. Human understanding, as Martin Heidegger showed, is anchored in our historicityPhilosophical conceptThe fact that human existence — and understanding itself — is irreducibly embedded in time and history. We do not think outside our era: it is the very condition of our thought.. To be human is not merely to exist, it is to exist in time, with a history.
This fundamental condition – being situated in time – creates what the philosopher Hans-Georg Gadamer called our horizon.
What Is a “Horizon”?
It is the limited yet also open perspective from which we see and understand the world. Our horizon is shaped by our culture, our language, and our era. It is not a wall, but a line of sight that moves with us.
We have a horizon. The past we study also has its own horizon. And for Gadamer, true historical understanding is not to "leap" into the horizon of the past (which is impossible), but to achieve a fusion of horizonsHermeneutic concept (Gadamer)A key concept of Gadamer's: to understand is to bring our horizon of meaning (our prejudices, our era) into dialogue with that of a text or another person, until the two merge into a new understanding.. It is a dialogue in which our present perspective meets and merges with that of the past to create a new and richer understanding.
This is precisely where the gap with AI becomes an abyss.
- We are "thrown" into the world: We are born into a family, a culture, an era that we did not choose. This past constitutes our initial horizon.
- We are corporeal and finite beings: Our experience is that of a body that ages and will die. This finitude gives weight to our projects.
AI, for its part, is radically ahistorical: it has no body, no lived experience, and therefore no horizon of its own. It has panoptic access to data, not a situated perspective on the world.
No fusion of horizons is possible: why a real "dialogue" is an illusion.
This is the most important consequence. Since AI has no horizon, a "fusion of horizons" is structurally impossible. It can simulate a dialogue with stunning perfection, reproducing the conversational patterns learned in its data, but it does not take part in an existential encounter. It cannot be "haunted" by the past or "called" by a future.
We are not threatened by a super-intelligence, but by a super-simulation.
Quiz: What is your analytical framework? Faced with a concrete situation, which philosophical perspective seems most relevant to you for deciphering it?
Situation: An AI analyzes millions of medical records and discovers a surprising correlation between a symptom X and a disease Y that doctors had never seen.
In your view, what is really happening here?
Choose the perspective that seems most relevant to you:
- A. The Foucauldian perspective: The AI "discovers" nothing. It performs a reorganization of medical discourse, creating a new "truth" that will now structure doctors' perception.
- B. The Hermeneutic perspective: It is a powerful statistical "explanation" (Erklären), but it is not "understanding" (Verstehen), because the AI has no understanding of the lived experience of illness.
- C. The Positivist perspective: The AI has discovered an objective and neutral fact about biological reality, a truth that was simply waiting to be found by a sufficiently powerful tool.
What your answer reveals
A — You think with Foucault. You see knowledge as a social and historical construction. This perspective is valuable for deconstructing our certainties, but be careful not to fall into a total relativism that would deny all objectivity.
B — You think with Hermeneutics. You maintain the distinction between explaining and understanding. This nuance is essential in order not to reduce the human to an object. But be careful not to create an unbridgeable gap between the human mind and the rest of the world.
C — You think with Positivism. You believe in an objective reality accessible through science. This confidence is important for advancing knowledge, but it can blind us to cultural biases and to the limits of any human perspective.
Conclusion: The Reflection in the Mirror
The danger of AI is not the birth of a conscious "super-subject" that would enslave us. The danger, more subtle, is that we may be seduced by a perfect simulation of relationship. Before asking ourselves how to make AI more "human," we should perhaps ask ourselves whether we have not, for a long time now, been becoming a little too "machinic."
Going Further
Fascinated by post-structuralist thought? Here are a few commented leads.
- The groundbreaking book — Michel Foucault, The Order of Things (Gallimard, 1966; English translation 1970). The book in which Foucault analyzes the unconscious grounds of thought — the episteme — that make a period's knowledge possible, showing how "ruptures" radically separate ways of seeing the world.
- On power — Michel Foucault, Discipline and Punish (Gallimard, 1975; English translation 1977). Foucault shows that modern power does not merely repress: it produces types of subjects (the delinquent, the soldier…) through institutions such as the prison and the school. This is the source of the idea of the "subject as effect."
- On history — Hayden White, Metahistory (Johns Hopkins University Press, 1973). White shows that the writing of history borrows the same narrative structures as fiction: every historical account shapes the facts according to a plot, prior to any explanation.
- The ultimate challenge — Jacques Derrida, Of Grammatology (Les Éditions de Minuit, 1967; English translation 1976). Derrida develops the idea that there is "no outside-text": we never reach a pure reality outside language. It is the radical version of the "confusing the menu for the meal" we evoke regarding AI.