There's something dizzying about watching a language model write in three seconds an essay that would take a student of the khâgne (elite preparatory class) three hours to produce. It's not just a matter of speed. It's a question of meaning: if the machine does it better, faster, and without complaint, then what is the point of learning? What is the point of effort? What is the point of a diploma? These questions, long confined to the futuristic columns of tech magazines, are now sitting in lecture halls, teachers' lounges, and HR offices. Artificial intelligence doesn't just threaten jobs: it threatens the very meaning of human intellectual work.

This text is not an anti-technology pamphlet. It's a clear-eyed attempt to map what AI is actually doing to our education systems, our labor markets, and our social contract — to draw, without angelism or catastrophism, the necessary conclusions.

The Diploma in an Era of Accelerated Obsolescence

The French National Education system is based on an implicit contract centuries old: the State trains capable citizens, citizens offer their skills to society, and society remunerates these skills through the labor market. This contract is now fractured on all sides.

The first problem is structural: the time for training no longer matches the time for the obsolescence of knowledge. A student starting a Master's in Computer Science in 2024 is learning languages, frameworks, and paradigms that will be partially outdated by their graduation in 2026. Not because of a sudden revolution, but because the pace of innovation in AI is compressing technological cycles to unprecedented levels. Universities, designed for long cycles of knowledge capitalization, find themselves teaching the day before yesterday's news.

A Few Figures Summarizing the Shift

- 38% of European jobs exposed to partial automation by 2030 (OECD, 2023).

- 5 years: average lifespan of a specialized technical skill in 2026, compared to twelve years in 2000.

- 62% of French high school students report regular use of AI for homework (Ipsos, 2025).

The second problem is pedagogical. For a long time, schools have evaluated the ability to *reproduce*: memorizing a formula, reciting a date, solving a standard exercise. This is precisely what large language models do with formidable effectiveness. The result: students no longer make the effort to memorize since they can ask instantly. Teachers no longer know what to evaluate. Exams lose their legitimacy when a phone is enough to get a 17/20 on a philosophy essay.

It's not the students' fault: the very structure of evaluation is obsolete. But changing the structure means challenging seventy years of pedagogical engineering built around the vertical transmission of codified knowledge — a titanic task that neither teacher unions nor successive ministries have the means to undertake urgently.

The third problem is social. Access to AI is not equal. A Parisian high school student with a premium subscription to an AI assistant has a tutor available twenty-four hours a day, patient, encyclopedic, and multilingual. A student in a rural area or from a disadvantaged family does not have this access. AI risks widening educational inequalities rather than reducing them, contrary to the enthusiastic promises of EdTech. The tool that could democratize access to knowledge will become an additional competitive advantage for those who already have everything.

When the Signal of the Diploma Loses Its Meaning

A diploma is a signal. It tells a recruiter: this person has survived five years of academic selection, they can read complexity, they know how to work under pressure, they are capable of learning. This signal has value because it is costly to produce — in time, money, and cognitive effort.

AI does not destroy the signal directly. It devalues it indirectly, in two simultaneous ways.

First devaluation: the skills certified by the diploma are partially replaced. A law graduate knows how to draft contracts, analyze case law, build an argument. Conversational assistants do the same — less well on nuances, much faster on volume. If law firms need one lawyer where they used to hire five, the surplus graduates find themselves demoted. The diploma still proves a skill; but the skill is worth less.

Second devaluation: the rarity value of the diploma collapses. For decades, the massification of higher education has gradually diluted the signal of the baccalaureate, then the licence, then the master's degree. AI accelerates this movement by making accessible to anyone the synthesis, analysis, and writing skills that constituted the core added value of intellectual professions.

Two Plausible Scenarios for the End of the Decade

Scenario 2028. A student spends five years at Sciences Po to master political analysis and institutional communication. Upon graduation, consulting firms have reduced their junior staff by 40% thanks to specialized AI tools. Their starting salary is 25% lower than that of the 2020 cohort. Their student loan, however, has not decreased.

Scenario 2030. CAC 40 companies begin recruiting based on portfolios and demonstrable projects rather than diplomas. Grandes écoles negotiate their survival by repositioning themselves on training for irreplaceable skills: emotional leadership, intercultural negotiation, decision-making under radical uncertainty. The certification market explodes.

The question is not 'are studies still worth anything?' — they are. But their value is shifting. It is migrating from codifiable technical skills to relational, creative, and ethical skills. Yet, these are precisely the skills that the National Education system is least equipped to teach and certify. The gap between what the school produces and what the market values has never been wider.

The Wave of Silent Layoffs

Economists have been debating for ten years: will AI destroy more jobs than it creates? The debate may be ill-posed. The real question is not the net balance over thirty years — it's the speed of destruction and the social geography of the victims.

The industrial revolution of the 19th century massively destroyed agricultural and artisanal jobs. But it did so over several generations. The children of weavers became factory workers, the children of factory workers became office clerks. Society had time to adapt, painfully but organically. AI does not have this kind of patience.

Layoffs related to AI do not resemble traditional social plans. They are rarely announced as such. A consulting firm does not lay off fifty junior analysts: it does not replace them when they leave. A newsroom does not eliminate twenty journalists: it outsources the production of standard content and reduces its permanent staff with the next wave of departures. This is what American economists call *soft displacement*: a slow evaporation rather than a visible wave.

The orders of magnitude, however, are not silent: 14 million jobs threatened in Europe by 2030 according to the World Economic Forum (2024), -26% of job openings in corporate finance between 2022 and 2025 in major French banks, and a destruction rate of cognitive jobs estimated at three times higher than that of manual jobs since 2023.

Why White-Collar Workers Are Paying the Price

What makes this wave particularly destabilizing is that it primarily affects qualified middle classes — exactly the people who had followed the advice given for thirty years: 'get an education, work in the service sector, the white-collar job is safe.' The jobs least threatened in the short term paradoxically remain the non-routine manual jobs: plumber, electrician, caregiver. A robot doesn't lay a floor yet, nor does it comfort an elderly person. But a conversational assistant drafts contracts, prepares pitch decks, codes Python, translates documents, and answers customer emails.

Blue-collar workers had had their turn; now it's the white-collar workers'. The difference is that no one had warned them — and they had often bet everything on their diploma to escape it.

France, with its social model built on salaried employment, is particularly exposed. Labor law, social security contributions, pay-as-you-go pensions, unemployment insurance: the entire edifice rests on the assumption that almost all able-bodied adults work. If this assumption collapses, even partially, the entire system falters.

Universal Basic Income: The Big Unanswered Question

Universal Basic Income (UBI) is no longer a left-wing utopia from the 1970s. It is a hypothesis seriously discussed by economists of all stripes, from Silicon Valley CEOs to liberal European think tanks. The argument is simple: if AI permanently eliminates a significant fraction of available jobs, a classic market economy cannot absorb the shock. A universal safety net, decoupled from work, is needed.

But the simplicity of the argument hides formidable difficulties.

The funding problem. A truly universal UBI — paid to every French adult, unconditionally, at 900 euros per month — would cost about 480 billion euros per year. The French state budget is 600 billion. This cannot be financed by cutting a few tax loopholes: it involves a total overhaul of the tax system, likely massive taxation of capital, robots, and AI profits — a tax regime that the concerned economic actors would deploy all their energy to circumvent or offshore.

The meaning problem. Work is not just a means of subsistence. It is a source of identity, temporal structure, socialization, recognition. Pilot UBI experiments — in Finland, Kenya, California — show that beneficiaries live better materially but do not solve the question of meaning. A society where the majority of people have no productive economic role is a society we don't yet know how to organize.

The transition problem. Between the 'current world' and a 'world with UBI,' there is a phase of political, legal, and fiscal transition that could take a decade. However, job destruction by AI will not wait for governments to agree. The time lag between the speed of technological change and the speed of institutional change is perhaps the most underestimated risk.

Three Families, Three Answers

What progressives defend. A UBI financed by a tax on automation: each job replaced by AI would trigger a contribution paid to a national fund universally redistributed. Simple in theory, complex to measure and impose in open economies.

What liberals defend. Not a UBI but a Negative Income Tax (NIT): below a certain threshold, the state pays a decreasing benefit. Simpler to finance, but it leaves millions of people very low in the distribution.

What conservatives fear. The collapse of the work ethic, mass social disaffiliation, and increased dependence on the state that future generations would have to finance. Fears that deserve to be taken seriously without invalidating the analysis of the problem.

What is certain: the status quo is not sustainable. A society that allows the gap to widen between the wealth produced by AI systems and the living conditions of a growing fraction of its population is heading towards political instability that history has already named: populism, resentment, irreconcilable divisions.

Editorial Opinion

Artificial intelligence is an extraordinary tool. It can compress years of medical research, make legal services accessible that only the wealthy could afford, and multiply individuals' creative capacity. It is not the technology that is dangerous — it is the idea that it would deploy without society collectively deciding its rules.

What is at stake is the question of who benefits from the productivity gains generated by AI. If these gains are concentrated in the balance sheets of a few tech companies while millions of qualified workers lose their income and sense of purpose, then we will have achieved a remarkable technological feat while producing a social catastrophe.

Refusing to normalize this means demanding that education be rethought for what it truly is — not a factory for technical skills, but a school for critical, creative, and ethical humanity. It means demanding that the labor market be regulated in its transformation rather than abandoned to its logic. It means demanding that the debate on universal basic income be honestly addressed, without angelism or ideological rejection.

AI forces society to decide what it wants to be. Perhaps that is its most valuable — and most uncomfortable — contribution.

Key Takeaways

- 38% of European jobs exposed to partial automation by 2030 (OECD).

- 14 million jobs threatened in Europe by 2030 (World Economic Forum, 2024).

- Average lifespan of a specialized technical skill has fallen to 5 years in 2026, down from 12 in 2000.

- 62% of French high school students regularly use AI for their homework (Ipsos, 2025).

- A universal UBI at €900/month in France would cost ≈ €480 billion/year, or 80% of the state budget.

- Qualified middle-class service sector workers are now the most exposed to *soft displacement*.