In a People's Daily essay titled "In the Age of AI, What Should Education Do?", Shi Yigong argues that education today is still largely built on the productivity characteristics of the industrial era — centred on transmitting knowledge — and that AI's rise poses a serious challenge to that model, one he says goes deep into curriculum structure, not just classroom tools. His sharpest line, describing what he thinks higher education must stop doing, is a warning against manufacturing "knowledge containers" (知识容器) on an assembly line.

The industrial-era model, named plainly

It's worth being precise about what Shi Yigong is actually describing, because the phrase "industrial era" can sound like vague criticism rather than a specific claim. The model he's pointing at is one where education's central job is transferring a defined body of knowledge from teacher (or textbook) to student, as efficiently and reliably as possible, and testing how much of it the student retained. It's a genuinely reasonable design for a specific problem: producing a large, consistently-trained workforce at scale, when the knowledge required for most jobs was relatively stable and the main bottleneck was getting it into enough heads.

That model isn't a strawman — it's a real, historically successful design, and large parts of KS3 education in the UK, as elsewhere, still run on it: defined content, delivered on a schedule, assessed by how much of it a student can reproduce under exam conditions.

Why AI specifically breaks this bargain

Shi Yigong's argument for why this matters now, more urgently than it might have a decade ago, rests on a specific claim: AI "can effortlessly draw on almost all publicly available human knowledge and precisely take over standardised, repetitive labour." That's a direct statement about what AI is actually good at — not vague futurism, but a specific description of capability that maps closely onto what an education model built around knowledge transmission and recall was training students to be good at.

If a system whose core job is holding and retrieving broad knowledge, and executing standardised tasks reliably, now exists and is improving fast, then an education model whose core output is a student who's good at holding and retrieving knowledge and executing standardised tasks is training students to compete with something that's already better at exactly that. This is the specific mechanism behind Shi Yigong's warning, not a general anxiety about technology.

What he says must change, and why it's structural, not cosmetic

Shi Yigong is explicit that the response can't just be adding AI as a classroom tool on top of an otherwise unchanged model — he argues reform has to go into curriculum systems and knowledge structures themselves. That's a bigger claim than "use ChatGPT in lessons." It's a claim that the actual definition of what a well-educated student looks like needs to shift, because the old definition was implicitly calibrated against a world where humans held a near-monopoly on broad, retrievable knowledge — and that monopoly is gone.

What this means for a UK KS3 parent, specifically

The UK curriculum at KS3 is not, on the whole, structured around Shi Yigong's alternative — it remains substantially content-and-recall-oriented, for entirely reasonable institutional and assessment reasons that aren't going to change quickly. That gap is exactly why this argument is useful for a parent to hold onto even while the surrounding system doesn't fully reflect it yet: it's a reason not to treat "did they memorise this correctly" as the ceiling of what a KS3 education should be aiming for at home, alongside whatever the school itself is doing.

The other two articles in this series drawn from the same essay — on what AI can't do and on thinking beyond the standard answer — go into what Shi Yigong proposes education should actually optimise for instead. This one is about naming, clearly, what's no longer working: an education whose primary output is a well-stocked, reliably-retrievable knowledge container, competing directly against a technology that already does that better.

FAQ

What did Shi Yigong say about education and the industrial era?

He argues current education is built on industrial-era productivity assumptions, centred on transmitting knowledge, and that AI's rise challenges this model deeply — particularly in higher education, which he says must not simply manufacture "knowledge containers" on an assembly line.

Why does Shi Yigong think AI exposes this problem specifically now?

Because AI can now draw on almost all publicly available human knowledge and take over standardised, repetitive tasks with precision — an education model built around transmitting and testing recall of knowledge optimises students for exactly what AI already does better.

Does this mean KS3 subject knowledge doesn't matter anymore?

No — knowledge remains necessary raw material. The argument is that transmitting and testing recall of it can no longer be education's main job, because AI already does that job. What changes is what education needs to build on top of that knowledge.

Source

Adapted from 施一公 (Shi Yigong), "《人工智能时代，教育何为》" ("In the Age of AI, What Should Education Do?"), People's Daily, "大家手笔" column, 3 June 2026.

Duke Harewood · aitutors.me · Updated 14 Jul 2026.