The question no parent dares to ask out loud
You watch your child do their homework. Maths, conjugation, history and geography. You see their crossings-out, their hesitations, sometimes their tears over an exercise they cannot understand. And in your pocket, on your phone, there is an app that would solve all of it in 3 seconds. Better than you. Better than their teacher. Better than the best private tutors.
A thought crosses your mind, brief, unpleasant. You chase it away. But it comes back, like a pebble in your shoe. What's the point? What's the point of making this child cry over a Euclidean division (division with a remainder, taught in Year 4) that ChatGPT calculates in the blink of an eye? What's the point of making them memorise history dates that any voice assistant can recite? What's the point of any of it?
Few parents dare to ask this question out loud. Even fewer teachers. And yet it is probably the most important one of our time. Because for the first time in human history, we are passing on to our children skills that a machine already performs better than we do. Reading quickly, calculating accurately, writing correctly, memorising facts, speaking foreign languages: on all these ground that schools have structured for 200 years, the machine wins. And it will win more and more.
So what do we actually pass on? This article is an attempt at an honest answer.
Schools have not always taught the same things
To answer this question, we need some historical perspective. School as we know it is very recent: barely 150 years in France, with the Jules Ferry laws of 1881-1882 (Jules Ferry, French politician who made schooling free, secular and compulsory). Before that, the vast majority of children did not go to school. And the schools that did exist did not teach at all the same things as today.
In the Middle Ages, school mainly taught Latin, theology, rhetoric. Arithmetic? Useless, everyone used pebbles (calculus, in Latin, means "small pebble", which is where the word "calculation" comes from). Writing? Reserved for clerics. Reading? Often out loud, and collectively.
In the 19th century, with industrialisation, schools were completely reconfigured to produce disciplined workers and employees capable of obeying schedules, reading instructions, performing standardised operations. This is what the philosopher Michel Foucault called the "disciplinary" school: its purpose was as much to shape minds as to mould bodies.
In the 20th century, science, foreign languages and computing were added. Schools gradually adapted to the knowledge economy.
Schools have therefore never been fixed. They have always adapted to the tools and economic needs of each era. So the question is not "should schools change", but "how". And to answer that, we need to understand what really changes with AI.
When calculators arrived en masse in schools in the 1980s, the French debate was fierce. Should they be banned? Should they be allowed? Many teachers predicted the end of mental arithmetic, the end of the ability to count. Forty years later, mental arithmetic still exists, but its nature has changed. We learn less to set out a division than to estimate an order of magnitude. This is probably the right model for thinking about the arrival of AI: not an end, but a transformation.
What the machine does not do (and what becomes precious)
To answer "what do we pass on", we need to name precisely what AI cannot do. Here is what we observe in 2026.
AI desires nothing. It wants nothing. It does not wake up in the morning wondering what would be meaningful in its life. Desire, intrinsic motivation, what drives a human to get up and do something rather than nothing: that is strictly human. AI carries out requests, but does not formulate its own intention. Yet it is desire that structures everything: what we choose to learn, what we seek, what we become.
AI has no body. It does not know what it is to run until you are out of breath, to touch skin, to eat a ripe fruit picked from the tree, to catch cold, to have aching legs after a hike, to feel exhaustion after a sleepless night. This bodily experience profoundly structures human intelligence. Modern neuroscience has confirmed it: we do not think only with our brain, we think with our whole body. This is what philosophers call embodied cognition.
AI does not really make mistakes. It sometimes hallucinates (produces false information), but it is not afraid of being wrong. It does not suffer from failure, does not feel the shame of having said something stupid. Yet it is probably our relationship to error that makes us most deeply human. The desire to understand often comes from having been wrong and having suffered for it.
AI does not die. It does not know what that dull urgency that structures all human life is: we have only a few decades. This finitude shapes our choices, our priorities, our loves. AI is eternal (or nearly so) and unchanging. It does not go through childhood, adolescence, maturity, old age. It does not lose its parents. It does not become a parent. It does not wonder what it will leave behind.
AI is not alone with anyone. It speaks to millions of people simultaneously, without particularly loving any of them. Yet exclusive attention, time given without calculation, the patient presence of one human for another: that is one of the rarest and most precious goods. A parent who truly listens to their child for 20 minutes gives them something no AI can give, even the most capable one.
What children still have to learn
From these observations, we can begin to sketch out what we must absolutely continue to pass on. Not out of nostalgia, but out of necessity.
Learning to want. Helping a child discover what truly excites them, what they are drawn towards, what they love so much that they get up early to do it. This skill is not taught in any school curriculum. It is discovered over the long term, through observation, through the right to be bored, through the space to dream. At a time of omnipresent screens and constant demands on our attention, this may be the most precious thing to protect.
Learning to inhabit your body. Sport, nature, cooking, craft, music played on a real instrument: everything that anchors us in physical reality. An entire generation risks growing up without ever having planted a seed, climbed a tree, or made bread. These skills are not archaic: they are existential counterweights to an increasingly virtual life. Parents who pass these on give their children an anchor that AI cannot dematerialise.
Learning to wait, to persevere, to start again. Instant gratification (having everything, right now) is the enemy of deep learning. Yet AI accelerates this logic even further: you ask, you get. If you do not teach a child that it sometimes takes three years to become good at something, and that they must accept being bad at it at first, you remove any possibility of real mastery. Patience is a rare skill. It is learned very early or not at all.
Learning to judge. AI proposes, humans dispose. But to dispose properly, you need to be able to judge. And judging requires knowledge, context, experience, nuance. A child who cannot tell good information from bad, a good intuition from a bad one, a good friend from a bad one, is completely vulnerable to the proliferation of AI content. As we explored in our article on deepfakes and our analysis of faked demos, being able to tell truth from falsehood has become a vital skill.
Learning to love. Loving real people, in their imperfection, their fatigue, their changeability. Loving complex works that demand time. Loving places you have come to know. Loving a craft you have delved into. Exclusive love (as opposed to the love-service provided by AI companions) requires patience and effort. As we explored in our article on the economics of AI companions, many young people risk growing up with a diminished understanding of what human love is.
The more capable AI becomes, the more the skills we consider "basic" turn out to be valuable. Being able to read a book slowly and in depth. Being able to hold a two-hour conversation without a phone. Being able to cook a meal for ten people. Being able to plant something and make it grow. Being able to stay silent and listen to someone. These simple human gestures become, paradoxically, distinctive skills in a world where AI does the rest.
The school of tomorrow: three concrete avenues
Without lapsing into pedagogical activism, we can sketch a few transformations that French and European schools will probably have to adopt over the next ten years.
Avenue 1: fewer subjects, more depth. The current French curriculum is encyclopaedic: pupils learn to skim over a great many topics. At a time when AI can skim infinitely better than a human, that loses its point. Schools will probably have to choose: few subjects, but covered in real depth, with time allowed. Finland began this transition in the 2010s. It has since become one of the most highly regarded education systems in the world, according to international comparisons such as PISA (Programme for International Student Assessment, an OECD ranking).
Avenue 2: more cooperation, less competition. French schools are built around individual competition: marks, rankings, competitive exams. That logic made sense in an economy that sought to sort out the best individuals. In an economy augmented by AI, value shifts towards collective capabilities: being able to cooperate, communicate, build together. Several pedagogical approaches (Freinet, Montessori, democratic schools) have been saying this for a century. They are finally becoming relevant on a mass scale.
Avenue 3: more practice, less theory. AI excels at theory. It is weak on embodied practice. Schools will probably have to shift their centre of gravity towards experimentation, hands-on projects, tinkering, fieldwork. As we explored in our article on AI in education, some teachers are already making this pivot, but without any institutional framework.
The changing role of parents
Beyond school, parents carry a growing share of what gets passed on. Here is what looks set to change.
The parent as protector of long timeframes. Children are already saturated with fast stimulation: YouTube Shorts, TikTok, video games, notifications. The parent who protects stretches of slow time (reading, conversation, walking, creative boredom) gives their child an invaluable gift. Silence and boredom are not voids; they are the conditions for deep thought and creativity.
The parent as transmitter of stories. An AI can tell any story. But it cannot tell your stories: those of your family, your ancestors, your place, your era. This oral transmission, which profoundly shapes identity, is what AI will never be able to do in your place. And it is probably what today's generations lack most.
The parent as living model. Children learn by imitating. If their parents spend their evenings on their phones, so will they. If their parents read, cook, go for walks, talk to their friends in person, so will they. AI can advise, but it cannot be a living model. Parents who take on this embodied responsibility are probably doing the most important work of our time.
The parent as transmitter of the meaning of work. As we explored in our article on the disruption of professions and our analysis of the transformation of journalism, professions are going to be transformed massively. Preparing a child for a specific job is becoming absurd: we do not know which ones will exist in 15 years. But passing on the meaning of work (the joy of doing things well, pride in the result, cooperation with colleagues) is timeless.
The risk that must also be named
One could end this article on a note of cautious optimism. That would be incomplete. Several risks are real and deserve to be named.
The risk of a cognitive divide. Children from educated families will learn to use AI intelligently, to keep their critical faculties, to maintain their capacity for attention. Others risk being absorbed by algorithms, losing their ability to concentrate, suffering AI rather than mastering it. This divide could be worse than the digital divide of the 2000s.
The risk of linguistic impoverishment. If children ask ChatGPT to "summarise" all their books instead of reading them, they will not develop the vocabulary, the complex syntax, the capacity for argument that deep reading builds. As several recent studies have documented, the lexical level of French secondary-school pupils has already declined. AI can accelerate this trend or reverse it, depending on how it is integrated.
The risk of emotional dependence. As we documented in our article on AI companions, 72% of American teenagers have already tried an AI companion. If a child grows up discussing their doubts with an AI rather than with their parents or friends, they will develop a certain social competence less. The difficult conversation with a human cannot be simulated.
The civilisational risk. If an entire generation grows up systematically delegating its thinking to tools, the whole of civilisation loses cognitive skills accumulated over centuries. This is what the philosopher Bernard Stiegler called the "proletarianisation" of knowledge: the gradual loss of human capacities in favour of technical tools. The phenomenon is real and documented since the Neolithic: writing made us lose oral memory, printing transformed cultural memory, the internet externalised part of our factual knowledge. AI could be the next stage of this dispossession, or on the contrary the occasion for a renaissance, depending on collective choices.
What we really pass on
If it had to be summed up in one sentence, here is what we pass on to children in the age of AI. Not content, but a relationship to the world.
A relationship to time: learning to wait, to dig deeper, to start again. A relationship to others: loving, cooperating, listening. A relationship to oneself: knowing oneself, desiring, choosing. A relationship to the real: touching, creating, transforming.
AI does a great many things. It does not do that. And that is probably what means human transmission will remain necessary, even when machines do everything else.
One last thought. You are watching your child do their homework. You see their crossings-out, their hesitations, sometimes their tears. And you now know that it is not so much long division that matters. It is the fact that you are there, beside them, at that moment. That you watch them struggle. That you help them back up. That you show them, by your simple presence, that difficulty is part of life, and that they are capable of getting through it.
That presence is what we pass on. And it is probably what future generations will look back on as the most precious thing. To go further, you can read our analysis of the 1, 5 and 30-year scenarios, which projects these issues over time, and our article on scientific discoveries by AI, which questions the human place in the face of machines that discover.