In his People's Daily essay on AI and education, Shi Yigong argues that real-world problems can rarely be understood or solved using knowledge from a single discipline — and that breakthrough innovation in science itself often happens at the intersection of fields. His specific proposal is to use AI as a general-purpose connective tool: something that links knowledge, perspectives, and methods across disciplines, creating genuinely multidisciplinary learning environments that put students at what he calls the "front line" of innovation and creation, building their capacity to handle complex problems that don't respect subject boundaries.

Why this is a different argument from "cross-disciplinary research is good"

A related article in this series covers Shi Yigong's argument, from his basic-research essay, that elite institutions like Westlake University deliberately build cross-group collaboration into how they organise research. This is a related but distinct claim, from his AI-and-education essay: that AI specifically can make multidisciplinary connection achievable at a much larger scale than specially resourced research institutions or bespoke cross-curricular school units, which most schools simply can't build and staff regularly.

That's a meaningfully different, more optimistic claim. Building a genuine cross-disciplinary research culture, as Westlake does, requires significant institutional investment — joint appointments, restructured incentives, dedicated space for collaboration. Shi Yigong's argument about AI in education is that a general-purpose tool capable of connecting knowledge across fields on demand doesn't require that same scale of institutional rebuilding to start having an effect.

Why subject silos are a genuine, structural problem, not just an inconvenience

The UK KS3 curriculum, like most secondary curricula, is organised into separate subjects with separate teachers, separate rooms, separate specifications, and — practically — very little built-in mechanism for a student to be shown, systematically, how a concept in one subject shows up in another. This isn't a design flaw exactly; subject specialisation lets teachers go deep and lets assessment be consistent. But it does mean that, left unaddressed, a student can go through an entire year genuinely not noticing that the exponential growth curve in maths is the same shape as population growth in biology, or that a geographic feature they studied in one term explains a historical event they studied in another.

Shi Yigong's underlying claim, echoing the research-institution argument, is that this kind of connection-noticing isn't a nice bonus — it's close to the actual mechanism behind original thinking. A student trained entirely within sealed subject compartments is being trained toward exactly the opposite of the capacity he thinks matters most.

Where AI genuinely changes what's practically achievable

This is the part of the argument worth taking seriously rather than dismissing as generic AI-optimism. Building a bespoke cross-curricular unit connecting, say, chemistry and history requires a teacher (or two) to design something outside their normal specification, with no guarantee any given student needs that particular connection at that particular moment. A tool that can, on demand, connect whatever a student is currently working on in one subject to its expression in another — when a student is curious enough to ask, or when a tutor is designed to make the connection visible — doesn't require that upfront institutional investment. It makes the connection available at the moment a student is actually engaged with the material, rather than only on the occasional day a special cross-curricular unit happens to be scheduled.

What good use of this actually looks like, versus a gimmick

It's worth distinguishing genuine cross-disciplinary connection from a superficial one. A tutor that occasionally drops in an unrelated "fun fact" from another subject isn't doing what Shi Yigong describes — that's decoration, not the "front line of innovation and creation" he's actually pointing at. What he's describing is closer to: a student working on a maths concept being shown, at the moment it's relevant, how the same underlying structure explains something they're separately studying in science or geography, in a way that changes how they understand both. That requires the connection to be substantive and well-timed, not just present.

FAQ

What does Shi Yigong mean by 'tearing down disciplinary silos'?

He argues real-world problems can rarely be solved using knowledge from a single discipline, and that scientific breakthroughs often happen at the intersection of fields. He proposes using AI as a general-purpose tool to connect knowledge, perspectives, and methods across disciplines for students.

Is this the same as the argument for crossing disciplines in research?

It's closely related but distinct — the research argument is about how breakthroughs happen at elite institutions like Westlake. This part of his AI-and-education essay is about using AI to make multidisciplinary connection achievable inside ordinary education more broadly.

How can AI actually help tear down subject silos at KS3 level?

By making it easier to connect a concept in one subject to its expression in another — on demand, at the moment a student is engaged — rather than requiring a specially designed cross-curricular unit most schools can't build and staff regularly.

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.