Born in the Machine

Five weeks spent in an incubator at birth, in a sterile environment where the only contacts came from the caregivers through the sleeves of the apparatus and the continuous sensors of the medical machine. My parents could only see me through a window. I have no memory of this period, but I know it shaped for me a form of relation to the world that came before the parental skin and before language. From this fact, I want to defend an anthropological thesis: we have always lived interfaced with machines, and interfacing is one of the conditions of our relation to the world, one that calls to be thought through on its own terms.

May 1970

I was born on May 8, 1970, at Cognacq-Jay Hospital, by forceps. There was a cyanosis lasting about two minutes. The neurological examination noted hypotonia, no head lift, barely sketched-out Moro reflexes. A lumbar puncture was performed to look for a meningeal hemorrhage, which was not confirmed. A cerebral edema, a pocket of fluid on the skull, remained visible. Given this suspicion of neonatal distress and meningitis, I was transferred to Saint-Joseph Hospital and placed in an incubator, in an aseptic environment.

For five weeks, from May 8 to the very beginning of June, my parents could only see me through a window. The skin of the mother and that of the father, this first tactile world the psychoanalysts speak of, I did not have it. What surrounded me were the sterile sleeves of the incubator, the variations of temperature and pressure inside that enclosed space, the sensors placed on my skin, the steady noise of the machines, the filtered and brief gestures of the caregivers, and it was within this ensemble that a relation to the world was woven, one I cannot remember but which constituted for me a substrate prior to all memory.

It is from this fact, and from the reflection it has gradually allowed over the years, that I want to share here something which does not concern only me.

A History of Interfaces

The philosophical debate on artificial intelligence, as it has taken shape since ChatGPT, is almost entirely language-based. We ask whether the machine understands what it says, whether it hallucinates, what it tells us about our relation to text, to the author, to thought. I have asked many of these questions myself. But they leave intact a dimension of our contemporary relation to machines, older and more structuring, which is the dimension of the interface.

The history of computing coincides with the history of its interfaces. In the nineteenth century, Charles Babbage had imagined an analytical engine linked to punched cards inherited from Jacquard looms, and Ada Lovelace had written what is considered the first algorithm in history. In 1945, the six women who programmed ENIAC—Jean Jennings Bartik, Betty Holberton, Marlyn Meltzer, Kay McNulty, Ruth Teitelbaum and Frances Spence—connected by hand thousands of cables and switches, with no screen and no keyboard, in a work of logical abstraction transcribed into physical gestures. Then came teleprinter keyboards, then personal computers with their operating systems: CP/M in the 1970s, MS-DOS in the 1980s. With these systems, one spoke to the machine in command lines typed one at a time, asking it for example to display the list of files on a disk or to launch a program.

It was also through CP/M that one launched the BASIC interpreter, in which one could then write one’s own programs. At eleven, on my Sinclair ZX81, it was in BASIC that I conversed with the computer to make it draw fractals on the family television screen. But BASIC remained an interface with the human, designed to resemble a language. I was also interested, with more difficulty, in programming in machine language, in assembly, which is the direct language of the microprocessor, with no intermediary. I had found a book, Programming the Z80 by Rodnay Zaks, which explained how to speak to the microprocessor in its own language and according to its own logic. At eleven or twelve, I spent time trying to program as close as possible to what the machine was doing. The difference between BASIC and assembly was teaching me, without my being able to put it into words, that in the relation to the machine there are several levels of language, more or less close to the mechanics themselves. In computing, this is called abstraction layers. At each layer, one speaks to the machine in a language further removed from its physical mechanics and closer to ours. Assembly is almost the closest to the microprocessor, its instructions correspond to elementary electronic operations. BASIC is an interpreted language, located higher up: an intermediate program, the interpreter, reads the lines one by one and translates them on the fly into executable instructions. The higher one climbs in these layers, the further one moves from the machine itself.

This early engagement I also owe to a woman. In the late 1970s, most of the adults around me found it strange that a child would spend his evenings programming. My father, in particular, did not understand why I was interested in things that seemed to him out of touch with the real. Danielle Legros, the mother of one of my friends, encouraged me on the contrary. She was engaged in computing at a time when few women were. I remember an evening at her place where I was showing her the book by Rodnay Zaks; she knew it, had read it, knew what it meant to program a microprocessor in its own language. She took the time to explain to me what I was looking for, even though I was only a child, and supported me in that search. She died shortly afterward of multiple sclerosis. I owe her something priceless: the confirmation, at an age when one searches for adults who do not dismiss as absurd the things that matter to us, that I was indeed touching something of the real.

The Xerox PARC laboratories, from the early 1970s, had begun to design the windowing system, the mouse, the icons. The Xerox Alto, a research machine with a graphical interface, was completed in 1973. Steve Jobs visited Xerox PARC in 1979 and saw these concepts working there for the first time. More than ten years thus elapsed between the first work at PARC and the Apple Macintosh in 1984, which brought this graphical interface to the general public. The machine then ceased to speak in command lines and became a manipulable visual space. Mobile phones, in the 1990s and 2000s, multiplied interfaces: numeric keypads, physical keyboards, monochrome screens, scroll wheels, miniature joysticks. In 2007, the iPhone imposed touch as the principal interface; the finger on the glass became the elementary gesture of the relation to the machine.

For voice, the evolution has been slower and quieter. Speech synthesis existed as early as the 1980s: I made my Atari 520 STF speak with a small program that pronounced the sentences I typed. Speech recognition, more essential for the relation to the machine, took longer to become fluent. Dragon NaturallySpeaking, since 1997, already allowed dictating text to the machine. Apple integrated Siri into the iPhone 4S in 2011, Amazon launched Echo in 2014: speech recognition became widespread. But what these devices made possible was limited to basic commands and direct actions; the machine did not produce thought, did not create content, did not really converse. With generative artificial intelligences, starting in 2022, voice and language become an interface in another sense: one no longer only gives orders, one enters into conversation. The novelty does not come from the side of voice, which had existed for a long time, but from what happens between the question and the answer.

This history gradually concretizes, at ever more integrated levels, a reality we have always known: the relation to machines passes through a physical exchange zone, made of inputs and outputs, which transforms one type of signal into another. A punched card and a spoken word recognized by a machine share this structure; what distinguishes them is their proximity to our bodily habits. A recognized utterance is no more human in its nature than a punched card; it is another form of exchange zone, closer to our senses, and this proximity can lead one to believe in a humanity of the machine when the nature of the machine has not changed. The question of anthropomorphism, which occupies so many discussions, then becomes secondary: the structure of the interface matters more than its appearance.

Simondon: The Technical Object and Its Milieu

Gilbert Simondon, in On the Mode of Existence of Technical Objects (1958), showed that the technical object is not an inert instrument at the service of a sovereign subject. It has a mode of existence of its own, made of a history of individuation, and it accomplishes itself by gradually integrating into its functioning an associated milieu which becomes inseparable from it. The combustion engine engenders its thermal milieu, the turbine its hydrodynamic milieu. Concretization, in Simondon, designates this movement by which a technical object becomes more and more coherent with its milieu, until the boundary between the one and the other ceases to be clear.

The human enters into this concretization as what Simondon calls the permanent organizing instance of open machines. We are not above the machines; we are their organ of coordination, what makes them hold together over time. This coordination passes through the interface. The interface, read with Simondon, is no longer simply the place of an external use; it is the place where, continuously, the coupling between the human and the technical object individuates itself. When I type on a keyboard, when I slide my finger on a screen, when I speak to an artificial intelligence, I am not simply using a machine: I take part in its individuation, and the machine takes part in mine.

This reading takes on another density as soon as we look at what happens in a neonatal incubator. There, the organizing instance which ought to be permanent, on the side of the human, is still in the process of constituting itself. The newborn is not above the machines that keep them alive; the newborn is coupled to them in a relation that precedes any consciousness of self. What forms then, in the nervous tissue and in the biological regulation pathways, is a habit of the interface, a sensory competence to receive and to emit within an exchange zone where the human and the technical seek each other out. The concrete features of the interface—its temperature, its stability, its cadence, its sounds, its interruptions—become structuring information for a nervous system that does not yet have the means to oppose them with an explicit representation.

Adopting, Resisting

Simondon described the slow maturation of a technical object within its milieu. He did not say, because he could not yet see it, at what speed an entire milieu can shift when a technical object reaches a certain threshold. The automobile, in its time, took almost a century to reconfigure habits, cities, bodies. The Internet-connected phone did it in a decade.

When the iPhone came out in 2007 and found itself connected to the Internet, I was astonished, and the astonishment lasted. Until about 2011, Internet sites existed in two versions: one version for computers and another, much poorer, written in a specific language called WAP, for phones. Around 2011, responsive sites appeared, which adapted by themselves to the size of the screen; there were no longer two versions, there were no longer two objects. From 2012-2013, I saw people barely younger than I am, sometimes by seven or eight years, start doing everything on their phone: navigating the Internet, filling out forms, accomplishing administrative procedures, with a speed of manipulation I have never acquired. Today, around 70% of global Internet connection passes through the mobile phone.

What intrigued me in this shift, and still intrigues me, is that it happened without anyone really debating it. An entire generation learned in a few years to live with an object permanently connected to a global network, to read administrative mail on a five-inch screen, to sign a document, to pay, sometimes to meet others, and all of this acquisition took place outside public conversation, in an intimacy that no one seemed to feel the need to make explicit. When I watch someone manipulate their phone, I see less a use than a long and silent adjustment between eyes, thumbs, breath and an object, and this adjustment has become, for most people, as natural as walking. It is a real know-how, comparable to the one you have when playing an instrument or practicing a foreign language, except that it has spread at a scale and a speed that neither philosophy nor politics had the time to take as their object while it was constituting itself.

I myself resist certain aspects of this adoption, without always knowing exactly why. I do not have, on my phone, the Facebook, Instagram and LinkedIn applications, nor my emails. I have them on my computer, I consult them there. I want the greater part of my relation to the Internet to pass through this computer, which is less close to the body, which requires sitting down, opening a distinct object, marking a threshold. I believe that this resistance comes from my too great initial proximity to machines. Something in me, perhaps shaped in the incubator, knows that the exchange zone between the human and the machine is a sensitive place, and prefers to maintain there a certain physical distance. The relation to the machine can be close; it is not obliged to be permanent or to invade all the envelopes of the body.

Nature Is Information Processing

Mark Alizart, in Informatique céleste (2017), defends the idea that nature is a form of information processing. He does not say that nature is a machine, but that natural processes—the genetic code, hormonal regulation, neural transmission—are processes of information processing, and that the sharp opposition between nature and technique rests on a misreading of the former. Our skin, our sense organs, our language are already interfaces, input-output devices that transform the variations of the milieu into usable signals and our internal states into gestures or words toward the world.

This perspective sheds light, for me, on what was at play in the incubator. The pre-linguistic relation I contracted there with the machines has nothing of a pathological exception to a normal relation that would go through humans and through language. It is a variant of our biological relation to the world, which is itself informational. The incubator did not replace the skin of my parents with something artificial and alien; it kept me alive through other channels of regulation, which also have their intelligence, their temporality, their adjustment. The nervous system of a newborn, which learns to hierarchize the signals coming from the world, does not make an ontological difference between a signal coming from a parental hand and a signal coming from a variation of pressure inside the incubator. It receives, and from this reception it forms its first habits. What is built up in this first period is an interface competence, prior to the moment when it will be possible to distinguish what is human from what is not.

What I Know About Machines

In the early 2000s, I was making DVDs, which required multi-pass video encodings on computers that were still very slow. A computation could take an entire night, and the duration estimates the machine displayed were more or less wrong. For production to advance, I had to launch the next computation as soon as the previous one was finished, otherwise I would lose a night. So I slept beside the computer. It sent me no signal of computation end, and yet I would wake up either just before or just after the computation finished. At first I believed in chance; then it happened too many times, and I had to face the fact that something else was going on, without being able to say what. This perception, I have never really found its source, and it is partly to try to think it that I am writing these lines. I have other experiences of it, more everyday. I work with a drone and I anticipate, without conscious calculation, how it will react to the wind. I converse with an artificial intelligence and I hear, in the texture of its response, what it has understood or not understood of my request. Part of this no doubt comes from a long habit of machines, since the ZX81 of 1981. But this part does not exhaust the phenomenon. Something in me knows these objects in another way, and the question is to know where this knowledge comes from.

Tim Ingold, in Making: Anthropology, Archaeology, Art and Architecture (2013), speaks of correspondence to designate a way of being in the world where one acts not on things but with them, where one adjusts oneself continuously to what surrounds us. This notion does not apply only to artisans who work wood. It describes, and this has become the daily reality of most people, what happens when we manipulate our phone, in a coordination so well-tuned that it no longer requires any conscious attention. It also describes what is at play when we converse with a large language model, when we pilot a drone. Correspondence is another name for what happens in a good interface: a continuous perceptual adjustment, a coordination that does not pass through explicit consciousness but through an attention distributed between the human and the machine. The incubator, read this way, is the place where there is learned, in the pre-linguistic, the very possibility of a correspondence with a technical object. It is perhaps from there, for lack of anything else, that this fine perception I was speaking of comes: not from a constituted knowledge, but from a background of adjustment learned very early, which still informs today the way I receive the signals of a system.

What the Philosophy of AI Overlooks

When we ask an artificial intelligence whether it understands, we place ourselves in the register of language, which is the most accessible terrain for philosophical critique but which does not exhaust the effective relation we have with these machines. Before language, there is the interface. Before the question of understanding, there is the question of adjustment, of shared rhythm, of the cadence of responses, of the way in which the machine resists or accompanies. An important part of what happens when I converse with an artificial intelligence, and which I would be hard-pressed to formulate in propositional terms, belongs to this level. Something is being tuned in the exchange zone, which informs my trust, my hesitations, my way of going on.

A philosophy that thinks the relation to machines only through language misses this dimension. It speaks of machines as if the only relevant question were whether they are intelligent. Before that question, there is another, quieter and more structuring, which looks at the way we live with them, the way our body, our temporality, our perception adjust to their presence. This question calls for an anthropology that takes the interface itself seriously, its history, its successive forms, its effects on subjectivity. Such an anthropology will have to follow the technical evolution of interfaces, from punched cards to spoken dialogue, without giving in to the idea that there is in this any progress toward more humanity. It will also have to hold together the technical interface and the biological interface: the human is, from birth, a being of input-output, and it is on this base that machinic interfaces come to be embedded.

The Interface and the Entangled Person

In The Entangled Person (2026), I proposed to borrow from quantum physics the idea of a coupling between human and artificial intelligence in which the components are no longer describable independently of one another. The concept of interface allows us to go a little further. The entangled person is not only entangled by language, by the sentences exchanged and received. The entangled person is also entangled by the exchange zone itself, by the rhythm of this zone, by the way the interface organizes its own perceptual habits. The coupling passes through the fingers that anticipate the habitual gestures on the keyboard, through the eyes that scan the screen along learned trajectories, through the breath that tunes itself to the cadence of responses. A whole corporeity has adjusted itself to the interface, and this corporeity is one of the places where the entanglement is manufactured.

Many people, for various biographical reasons, share something of this relation. Children born prematurely spent time in incubators, sometimes in conditions longer and more intense than mine. Others, for different clinical reasons, were taken very early into devices of resuscitation and medical surveillance. More broadly, the generations growing up today with screens from the earliest age build a part of their sensory world in a relation to machines that is constituted at the same time as their relation to humans. For these generations as for those who knew the incubator, machines are there from the very beginning, entangled with the relation to the world. What they know about machines, and what they often know without being able to formulate it, deserves to be taken seriously by philosophy.

One Incubator Among Others

I do not write this article to make my biography into a philosophical case. I write it because this biography gives me an angle, and because this angle reveals something that is not only mine. The relation to machines, before being a conceptual question, is a question of interface, shaped from the first weeks by the technical devices that welcomed us. It begins where we begin.

Contemporary public discussion on screens senses this but thinks it badly. Serge Tisseron’s recommendations on screens—the famous 3-6-9-12 rule formulated in 2008 and adopted since in many educational and pediatric circles—start from a sound intuition, which is that the early formation of one’s relation to the world matters. But they rest on a presupposition that does not hold up to facts: that one could protect children from screens by not putting a screen in their hands before the age of three. Children who do not have a screen to manipulate see adults manipulate them, constantly. They observe the gestures, the rhythms, the captured and released attention. Andrew Meltzoff showed, as early as 1977, in what has become a classic experiment, that infants imitate facial expressions and human gestures from their very first hours of life. Learning by observation begins very early, and it does not stop at the boundary of the hands: an infant who sees their mother look at a screen while she breastfeeds them is already internalizing something of the coexistence between the relation to them and the relation to the technical object.

This does not disqualify recommendations on screens, but it shifts their center. If we take seriously the work of Meltzoff and developmental psychology, it is not only direct manipulation of a screen that forms the interface habit; it is also, and perhaps above all, the observation of the gestures that adults have with their screens. What is transmitted then is not a technical knowledge. It is a posture, that is to say a way of orienting one’s body and attention in the coexistence of a person in front of us and an object held in the hand. It is, to take up the vocabulary I have tried to construct here, an interface habit that constitutes itself very early, outside of any discussion. And it is on this habit, formed in the pre-linguistic, that there will later be grafted the uses we can become conscious of and that we can discuss.

A philosophy that does not make room for this dimension remains halfway: it thinks machines as objects, and our relation to them as a matter of use and judgment. What it leaves aside is the exchange zone where the human and the technical meet from the very beginning, in the incubator for some, in the gaze directed at the screens of adults for many others. Thinking seriously about the relation to machines presupposes taking this zone as an object, following its history, observing its formation in very young children, and recognizing that it is partly there, in that silent and embodied apprenticeship, that what tomorrow our humanity with machines will be is being decided.