Norway is the first country in the world to announce an almost total restriction on the use of Generative Artificial Intelligence (GenAI) in primary school.
In the early years of secondary school, its use may appear cautiously and under teacher supervision; in upper secondary school, by contrast, students are expected to learn how to use GenAI responsibly, with university and work in mind.
The measure comes after a decline in educational outcomes, a ban on smartphones in classrooms, and a shift toward more physical books and fewer tablets. Norway’s education minister argues that before children learn how to ask a machine for answers, they must learn how to read, write, calculate, and think.
Ultimately, what Norway is saying is that a human being in the early stages of formation needs to understand what the world has to say — and be able to enter into dialogue with it — directly, without anyone translating it on their behalf. This is the problem of the universal translator.
The universal translator
I still remember the first time I entered a printing shop. It operated inside a house with a gabled roof, on a small neighborhood street on the outskirts of Buenos Aires. The smell of metal blended well with the scent of ink and paper, and with the wooden shelves that held piles of notebooks, pads, and books.
Watching those mechanical hammers move, striking in a steady rhythm, fascinated me.
To me, that reproduction was not very different from handwriting, except for its capacity to create a series: many identical units of the same content.
The method was analog, and I had not yet realized that this machine defined the era into which I had been born.
Over time, I learned that Walter Benjamin had been concerned with the capacity for technical reproduction in the previous century. The German philosopher explained very well how this new thing — the series — modified the relationship between the Human Being and the work of art, because it eroded its aura: that unique, unrepeatable, situated presence that linked a body that looks with an object that is there, condensing meaning into a single sign.
I returned to that printing shop more than once, and confirmed that its system still preserved a physical relationship with matter. It was technical, yes. It was mechanical. But it did not seem false, even though its existence indicated that we had already found a way to produce culture without the need for the human hand.
The era we inhabit today is marked by a new printing press: GenAI. Nicholas Negroponte describes very well how we began our journey toward it, when he defined the passage from analog to digital, or from atoms to bits.
Texts, images, sounds, calculations, maps, memories, and bonds could all be translated into zeros and ones. When we did that, we placed between the Human Being and reality a universal translator, a mediation that intervenes even when we think about ourselves: programming code.
At first it was static, rough, predictable. But it allowed us to manipulate it, improve it, and eventually allow programs to begin learning. We call learning any action that arises from a certain autonomy of the system — no mechanical machine learns anything.
Thanks to that translation, known as information processing, we built systems capable of writing, programming, simulating, classifying, conversing, and creating synthetic images.
Benjamin would have gone mad just thinking about it.
Human or artificial?
Again and again we try to grasp the human, to detect it, to define it. Well, on the face of the Earth, no one other than the Human Being can create machines — nor artificial entities. So, although computer programs distance us much more from the direct experience of the world than the printing press ever did, they remain human; paradoxically, uniquely human.
But then Norway comes along and says: fine, that is not the issue. The issue is that a child is not the same as an adult. And they are probably right.
According to pedagogues, humans do not learn the way machines do. Especially in childhood, when we have not yet incorporated the ability to read and write, we learn with the body, in relation with other bodies, in a shared physical space, with smells, flavors, textures, and by recognizing the gaze of another adult.
But the smart devices through which we access GenAI privilege sight and, to a lesser extent, hearing, while weakening and reducing — by exclusion — smell, touch, and taste, the senses through which we connect physically with the world. On screens we even have representations of ourselves, but from the standpoint of our cognitive apparatus, they are very poor.
Following Heidegger, to inhabit the world is the fundamental experience of the Subject. And that inhabiting is only possible with the body, which digital interfaces cut down and reduce.
Hence the Norwegian precaution, which could be understood as a form of care: preserving for us the unique opportunity to inhabit and Be the world in childhood, without artificial mediation, in order to form Subjects who are as complete and fulfilled as possible. If we succeed in forming a free Subject — what a paradox — that Subject will later decide which tools he or she considers appropriate to use.
But what would happen if that Subject never even had the chance to experience the world directly, at least once?
According to Charles Peirce, there are forms of knowledge that depend on the index: the trace, the footprint, the contact, the physical sign that something was there. Many cultures that did not seek to formalize their knowledge — because doing so meant separating themselves from Nature — knew how to read the weather, the earth, animals, gestures, silences, smells.
As children, perhaps we resemble those people of old, who knew their surroundings intuitively but did not symbolize them — who did not create maps because they did not need them. With GPS, we can go almost anywhere in the world, but we depend on that interface that translates the territory for us.
The symbols that GenAI reads and recreates function because they are gathered into a code: that is, a list of signs whose meaning we agree is valid within the culture we inhabit.
Every language is a code. Music is another. Numbers are, too. Programming languages as well. They serve us to express reality so that someone else may know it without experiencing it.
In that sense, GenAI is the code of codes, because it is capable of understanding all the systems just mentioned, along with all the others that are digitally accessible, and it can even invent new ones.
Its power is overwhelming, and when a well-formed adult gains access to it, that person’s possibilities expand.
And yet, it is too complex, too powerful, too abstract an artifact to place in the hands of a child. Childhood is the moment in which the primary relationship between body, soul, language, imagination, world, and peers is formed.
Could we use GenAI in primary school, but with meticulous teacher assistance?
It seems unlikely — not because of the limitations of the technology, but because of the problem of institutions.
At home, when my 11-year-old daughter and I sit down to do homework, we often ask a GenAI tool to help us. The quantity and quality of the explanations it offers us in mathematics are extraordinary, but it is my presence as a responsible adult, together with my interventions, that allows the proper triangulation: in the living room, my daughter, as a student, finds in me a figure similar, in this case, to her teacher, whose role is that of responsible authority.
As a father, I authorize the use of GenAI, teach her how to use it, and also how to distrust its answers, qualify them, compare them with other sources and alternatives. I am not getting the problem out of my way; I am doubling down on my participation.
At home, GenAI is used for specific tasks, and the prevailing criterion is mine, gradually making room for my daughter’s, as I help her build critical thinking in the face of every answer. It is difficult to replicate that with more than five children.
If Norway reached the conclusion that it had to ban GenAI in the classroom because its schools were flooded with it, it is also worth considering that no one can prohibit such use in homes, where human and affective supervision makes all the difference.
And it is also worth asking whether there, as in every other society, when children come home, there is at least one Other who, with affection, accompanies the child in learning. Current statistics suggest that the answer is no: childhood autonomy in developed countries leaves plenty of room for this universal translator to whisper into our children’s ears.