Deutsch - English - Español - Français - Italiano
The terminal workshop of thought is being automated in its turn. First instalment of a four-part series on what the machine is doing to the school — and on who will decide what is done with the human.
In April 2026, the editorial board of the Journal of East China Normal University (Educational Sciences edition) published a one-page announcement. It comes down to a few words: the journal will publish research papers whose “first author” is an artificial intelligence.
The journal in question sits at the centre of the academic field. Monthly, ranked at the top of China’s citation indexes, supervised by the Ministry of Education, it is the journal that manufactured, in July 2023, the founding moment of the Chinese debate on AI in education: a sixteen-article ChatGPT special issue, downloaded tens of thousands of times. Three years later, the same editorial board writes that it is time to “stop the sterile debate over whether to use AI and enter the game resolutely”. The question of 2023 is declared closed by the very body that had posed it.
In Shanghai, 1,177 authors have the machine sign
The facts, as the announcement reports them. In September 2025, the university’s Faculty of Education, the Shanghai Institute of Intelligent Education and ECNU’s intelligent-education laboratory — accredited as a Ministry of Education social-science laboratory — launched what they present as “the world’s first large-scale social experiment in AI as first author” (AI一作): a call for research papers with AI as “first author and creative subject”, on the themes of future learning, the future classroom, future teachers, future schools and future research. 1,177 authors responded, from all 34 Chinese provincial-level divisions and nine countries, including the USA, the United Kingdom and South Korea. In March 2026, a closing symposium was held in Shanghai; it released a report entitled “Beyond the Turing Test” and announced a special issue: the full report, accompanied by five “pioneering papers of human-machine co-creation”, selected from ten by the committee. Each paper will appear as a three-part unit — the text, the committee’s recommendation, and a note describing the writing process.
The announcement is a promotional, self-celebrating text (“world first”, “a resounding stress test”); the figures are the organisers’ own, and the report has not yet appeared. Texts generated entirely by AI systems had, moreover, already passed peer review here and there. What the Shanghai announcement establishes is of a different order: a central institution of the Chinese educational field, under ministerial supervision, organises the thing on the scale of the whole country, theorises it (the experiment is explicitly placed in the service of the “fifth paradigm” of research and of the “construction of the autonomous knowledge system” of Chinese pedagogy), and gives it a publication protocol and a vocabulary — the symposium’s workshops were entitled “from tool to symbiosis”, “from contract to co-governance”, and the text sets the horizon of a “carbon-silicon symbiosis” (碳硅共生). The direction of the movement is established; how far it has actually advanced remains to be verified — the observation we applied to factory robots, whose demonstrations always run ahead of real production rates, holds here without amendment.
The “first author” is the one who answers for the work; the machine answers for nothing
In academic convention, the first-author position designates the person who did the essential part of the work and who answers for it. The signature commits a reputation; it exposes its bearer to retraction, sanction, discredit. It is the mechanism by which the scientific community imputes knowledge to someone.
An AI can commit nothing of the kind. It has no reputation to lose, cannot be sanctioned, answers for nothing before anyone. Making it sign as first author amounts to testing whether the signature can be decoupled from responsibility — and the Shanghai experiment tests exactly that, without ever saying so. The publication protocol bears the trace: if every paper must appear flanked by a note on its writing process, it is because the signature no longer suffices to establish the provenance of the work. We shall return to this in the third episode, for this displacement of proof reaches far beyond the laboratory.
The school pyramid drew its meaning from its summit; the summit is being automated
Why does an event in an educational-sciences journal concern the school as a whole? Because the educational system is a pyramid in which each storey is justified by the storey above. Primary school prepares for secondary school, which prepares for higher education, which prepares for the doctorate, which prepares for the production of new knowledge. Sorting takes place along this chain through mechanisms of different kinds. In China, a single competitive examination, the gaokao. In France, written-and-oral competitive examinations — the École normale supérieure, which skims off the future researchers, or the agrégation, the elite teachers’ examination — and, upstream, Parcoursup’s algorithmic file-based matching, which selects, for the most part, without any new test [Author's note, following an exchange with Denis M. The verb selects is to be read in the sense that the algorithm executes the selection produced upstream by the programmes' admissions committees. The algorithm itself is a stable matching and does not rank candidates]. Yet all of them order trajectories by the same yardstick: distance to the summit. The pyramid serves other ends too, vocational training first among them; but its legitimacy flowed down from the summit: research trained those who trained the others, and academic excellence was defined, link by link, by distance to the summit.
In our article on the workerless factory, we described a third regime of automation, distinct from the graft and the clean slate: the software that empties offices with neither body nor assembly line, whose milestones go by the names of prospectuses 95 % machine-drafted at Goldman Sachs and shrunken analyst classes at McKinsey. AI一作 is the limit case of this regime: the laboratory is the terminal workshop, the one in whose name all the other storeys existed. And the machine’s advantages there are exactly the robot’s on the line: it works at night, does not tire, iterates without ego, and its cost keeps falling.
When the summit is automated, the cascade descends. The doctorate loses the outlet that justified it, higher education loses its own, and so on down to the primary classroom. The question “what is school for” ceases to be a speculation for philosophers of education: it follows mechanically from the amputation. Nobody chose to pose this question; it now poses itself, and from the top.
The factory sends the human out; the laboratory keeps him as guarantor
There remains, between the factory and the laboratory, a difference that the Shanghai announcement exposes without meaning to. On the assembly line, when the machine produces, the worker leaves the workshop — that is the very definition of automation. In the journal, the institution decrees that the human stays. The announcement’s final sentence affirms it with lyricism: human researchers “do not leave the stage”; guided by “AI for good”, they remain to “guard the depth of thought and the radiance of humanity”.
The machine produces; the human vouches. The researcher moves from the status of producer to that of guarantor — lending his capacity to answer to an entity that cannot answer. The arrangement has its coherence, and it recalls a division of labour already theorised in China for the teacher, summoned to cede to the machine his role as master-of-knowledge and fall back on that of master-of-persons. It nonetheless leaves open a question the announcement does not ask: how long does a guarantor who no longer produces keep his post?
What this series will establish
At each wave of automation, the same promise has been remade: the machine will take the toil, man will keep the thinking. The power loom took the weaver’s gesture, the tractor the ploughman’s, office software the copyist’s — and each time, the value of the human was moved one storey upward, towards more school, more diplomas, more abstraction. The promise held as long as there was a storey above. AI一作 signals that there is none left: the work of thought, the terminal refuge, is being automated like the rest. The question Paul Lafargue put to the factory in 1880 — what becomes of the hour the machine cuts away, and who captures it — now poses itself with no way out, and it is on the school that it lands. The series that opens here follows it there.
The second episode will examine the two loudly announced resets of the school, the USAmerican and the Chinese. In Texas, a network of private schools has removed the teachers and cut academics down to two hours a day; in Hangzhou, a “school of the future” instruments its pupils’ every gesture, from the exercise book to the canteen. We will establish the precise point at which each of these refoundations stops — and show that this point is the same on both sides.
The third will follow the proof. When the product no longer proves anything, because anyone — or anything — may have made it, evaluation moves back towards the process. Three unrelated devices are making this shift at the same moment: a pen that captures the pupil’s reasoning stroke by stroke, a voice agent that has the student defend his paper, a fabrication note required of the researcher. Three storeys, one and the same migration of proof — and, on both sides of the world, the new instrument pointed at an old objective.
The fourth will tie the whole together. A worker paid three dollars an hour teaches the machine; a schoolgirl teaches it for free with every stroke of her pen; at the end of the chain, the machine signs the research in pedagogy. The loop is closed, and a closed loop always belongs to someone. By whom will the human be educated, and to whom will what is made of him belong — two questions that neither Washington nor Beijing nor Paris poses, while all three are already answering them in deeds.
The Shanghai announcement swears that the human “does not leave the stage”: he stays on it to watch over “the depth of thought and the radiance of humanity”. It remains to be established in what role — and under what contract.
All our publications are free and Copyleft. They may be republished, in whole or in part, under following conditions:
→ keeping the title and content
(text, images, editor’s or translator’s notes, boxes)
→ crediting the author(s) and/or translator(s)
→ inserting a clickable link to the source.
Any monetization or commercial use of these publications is prohibited and will be subject to legal action.