When Knowledge Is No Longer Scarce, What Are Schools For?

AI GF / AI summary Key ideas, distilled by AI Generate · quick read Generate

Summary ready 100%

01

The author argues that schooling centred on standardised instruction will be replaced, while education and universities will endure. AI will support personalised knowledge acquisition, with real relationships, collaboration and practice remaining central to growing up. The essay advocates learning around problems, combining specialist depth with exploration across fields, and continually connecting study with action. It criticises academic incentives that favour publishable results and calls on universities to protect important, long-term research that may fail. Universities retain value as places that concentrate people, research resources and tacit knowledge. As access to knowledge becomes easier, judgement, creativity, trust and the ability to find worthwhile problems become the scarce resources.

I believe that the school as we know it will be replaced.

I do not mean that education will disappear, or that children should grow up indoors, staring at screens. Our need to learn, socialise, collaborate and explore the world will not vanish because AI has arrived.

I mean the system we have grown accustomed to: arriving at a building at a fixed time each day, sitting in a classroom with dozens of people of roughly the same age, and following one teacher through the same material, at the same pace, towards the same exams.

It became widespread because, with limited resources, we had no way to offer everyone a different education—not because it was best for every individual.

As I see it, AI is dismantling the premise on which this model rests. Once a better alternative exists, there is no reason to cling to the old way of organising things forever.

1. When one-to-one teaching is no longer a luxury

A fundamental limitation of the traditional classroom is that a teacher can rarely give everyone the attention they need.

Facing dozens of students, a teacher must choose a pace that broadly suits the majority. Some find it too slow; others struggle to keep up. One student needs an equation, another a picture. Some understand immediately; others need to approach the idea from several directions.

This is not necessarily a failure of care. A class cannot pause indefinitely for any one person, and a teacher’s time, energy and patience all have limits.

So education is organised around standard age groups, curricula, textbooks, exams and graduation dates. From this perspective, modern schooling is an educational assembly line: its priority is to educate large numbers of people with a finite supply of teachers and resources. Helping each person learn as well as possible often comes second to the efficiency of the system.

I believe AI will fundamentally change that.

Imagine a sufficiently mature and reliable AI tutor. It could keep track of what a student has learned, what they have forgotten, and where their understanding first broke down. If one explanation does not work, it could try another: introducing calculus through a story, probability through a game, or physics through an image. A question asked ten times could be explained ten times.

If such a system could support learning consistently over many years, it might come to understand more than a decade of a person’s intellectual development. A child from an ordinary family might gain access to the sustained individual attention that only a few could previously afford.

I am not claiming that today’s AI is already a perfect teacher. My conviction concerns the direction of travel: as these capabilities mature, excellent teaching will cease to be a scarce resource available only to a few.

Then we will have to ask again: why must learning happen on the same timetable, at the same pace, in the same room?

2. Learning can be personal; growing up still requires other people

None of this means that children should withdraw from one another.

Schools perform a crucial function: they introduce children to a small society. What we learn there is never just mathematics and English. It is also how to make friends, deal with conflict, get along with people we dislike, cooperate, compete, accept responsibility, and recognise that others will not always accommodate us.

These abilities are difficult to develop solely through conversation with an AI that is always available to respond. Real people misunderstand us and turn us down. They have needs of their own. They can also help us in ways we never anticipated.

The future I imagine therefore needs more meaningful life together, in person.

The places that support it might still be called schools. Or they might look more like a combination of a community centre, club, sports facility, laboratory and project workshop. Children could spend the morning learning mathematics, languages and science with AI, then the afternoon playing football, building robots, making films, performing, running experiments or completing a real project with others.

Adults would remain essential. As mentors, coaches, project leaders and community organisers, they could help children develop judgement, navigate relationships and keep going when things become difficult.

If I had to put this division of roles into one sentence, it would be: AI teaches the knowledge; the real world teaches us how to grow up.

Functions now bundled into a single timetable do not have to remain bundled together. Education should be reorganised around what people need.

3. Universities will need to explain their value anew

At university, the question becomes sharper.

One of a university’s great attractions has been its ability to bring knowledge, professors, laboratories and capable people together. For someone seeking a systematic introduction to a field, university has usually offered a clear and dependable route.

But as knowledge becomes easier to access and explain, the university’s advantage as a gateway to it will inevitably weaken.

A learner can use AI to work through probability, asset pricing, stochastic processes, market microstructure and machine learning. If a paper is incomprehensible, they can ask their way back to the underlying concepts. If their mathematics is insufficient, they can return to it. If they already understand something, they need not wait for an entire class to finish.

A learner still has to practise, think and repeatedly test their understanding. But there is no reason to confine that work to a uniform course schedule.

Universities will therefore need a clearer answer to this question: if people no longer have to come here to gain knowledge, what makes it worth several years of their lives?

One familiar answer is the people.

The social value of university is real. Part of it comes from selection: bringing together people of similar ages who have met an academic threshold and share an interest in certain subjects. That is far more efficient than meeting strangers at random.

Yet the selection is imperfect. Two people studying finance need not have similar interests or ambitions, or be working on similar problems.

For someone who already knows what they want to do, an industry conference, research community, startup, open-source project or company can sometimes offer more immediately relevant conversations. The people there may differ in age and background, but share something more significant: they are seriously engaged in work that matters to you.

I used to think this meant that our connections would shift from classmates to people in the same field. Eventually, I realised that this was still too narrow.

The most valuable encounters may happen around a shared problem, rather than within a shared profession.

4. Real problems do not respect academic boundaries

In my own experience of university in Britain, I have never particularly liked the relatively fixed boundaries between undergraduate subjects.

I study finance, so naturally most of my courses revolve around finance, accounting and economics. If I become interested in philosophy, computer science or another field, I cannot combine courses across departments as freely as I would ideally like.

This is my personal experience, not a claim about every British university, let alone universities elsewhere. Some already offer considerable flexibility.

But the experience has made me ask: why should someone be defined so early, at eighteen or nineteen, by a single subject?

The real world is not organised into academic departments.

Consider running an investment business. On paper, it belongs to finance. In practice, it also requires software, data, AI, law, tax, regulation and organisational management. Investing involves emotion, cognitive bias and group behaviour. It can even lead us into philosophy:

What is value? How should we understand risk? In the face of uncertainty, what can we really know? When should we trust a model? Why do we keep mistaking luck for skill?

These questions do not begin by declaring which faculty they belong to.

Building a self-driving car is much the same. Software, mechanics, chips, mathematics, design, business, law, even ethics and urban planning may all become parts of a single problem.

Universities need disciplines to organise knowledge; action in the world tends to organise itself around problems. The former asks what you study. The latter is more interested in what you are trying to accomplish.

Seen this way, “interdisciplinary” is an intriguing word. From a department’s perspective, a problem crosses several boundaries. From the world’s perspective, the problem was whole to begin with.

We separated the knowledge first, then found that we needed to connect it again.

Classification is useful. The trouble begins when we mistake a way of organising knowledge for the boundaries of the world itself.

5. Bring different people together around the same problem

The value of practical environments, then, is not simply that they introduce me to more people in finance.

An investor meets an engineer. A programmer meets a designer. An AI researcher discusses consciousness with a philosopher; someone in finance talks to a psychologist about the mistakes people repeatedly make in markets. These conversations may produce nothing immediate, yet change how both people understand a problem.

Companies, open-source communities and collaborative projects can allow such exchanges to continue. People bring different knowledge, experience and judgement to the same table because they want to make something work.

One understands technology, another law, another the industry. Someone else brings ten years of failure and experimentation. Looking at the same problem, they often see entirely different layers.

I believe sustained work in such an environment often does more to broaden a person’s thinking than adding a few more courses in the same subject. It teaches you that your knowledge can be deep and still be far from complete.

6. A good education should leave room to wander

I do not think specialist knowledge will lose its value. Quite the opposite: genuinely difficult problems will always demand depth. Enthusiasm for crossing disciplines cannot substitute for the years of mathematical training required to work in theoretical physics.

But depth should not mean spending a lifetime inside one boundary.

Entering an unfamiliar field used to be costly: finding textbooks and teachers, planning a sequence of study, and slowly making sense of unfamiliar concepts. If AI can help people identify and fill the gaps in what they already know, that journey may become easier.

Someone with a finance background who wants to understand neural networks need not necessarily take another degree in computer science. Engineers can study economics, founders psychology, and doctors programming.

A person’s knowledge might then be described by more than “a degree in finance”. They might know finance deeply, write some software, understand AI and have a longstanding interest in philosophy. They may have built a business, invested and traded, and learned some law and tax because a particular project required it.

Their knowledge would look more like a growing network than a course catalogue drawn up before they enrolled.

Today, many apparent detours are easily dismissed as inefficient. Why study philosophy if your degree is in finance? Why spend so much time on history if you study computing? Why leave physics to start a business?

Yet these detours can be how a person develops a distinctive way of seeing.

An individual advantage need not come from taking one more course than everyone else in the same subject. It may emerge when specialist knowledge is repeatedly reshaped by technology, history, philosophy and lived experience, forming connections that other people have not made.

To me, one of AI’s greatest contributions to education will be that we no longer have to accept the same combination of knowledge.

Problems give learning its direction, curiosity expands its boundaries, and practice tests the depth of our understanding.

7. Learning and practice need not be separate stages of life

The familiar sequence is to study, graduate, find work, and only then begin to encounter real problems.

But learning can also happen continuously through doing.

A seventeen-year-old interested in robotics could start with a little linear algebra, build a simple robot, and study control theory when a control problem arises. They could put their code on GitHub, meet developers from other countries, attend events and join a real project.

Partway through, they might discover that their probability is weak and return to it. Later, trying to sell a product might lead them towards supply chains, product design and fundraising.

The foundations still have to be learned, but now they connect to a real problem. You know why you need them, and why you cannot move forward without them.

Learning, making, failing, learning again, collaborating and creating can form a continuous cycle.

Our connections with other people would then come not only from being in the same class, but also from working on the same problem.

To me, this is the deeper change AI brings to education: it changes not just who explains the knowledge, but the entire sequence of learning.

8. What kinds of questions does academic research reward?

Beyond teaching, I have another question about universities: as research becomes more specialised, how do we ensure that worthwhile questions are not squeezed out by the way work is assessed?

Narrow research is not inherently a bad thing. As science advances, it requires greater specialisation. The fact that a paper in particle physics or finance is inaccessible to a general reader does not establish that it lacks value.

What troubles me is that academic incentives seem too concerned with what is publishable, and not concerned enough with what is important.

Many influential strands of financial research began with questions that are straightforward to state: how should risk be priced? Are markets efficient? Why do firms choose debt rather than equity? Are investors rational? How should derivatives be valued?

We do not need to understand a long list of technical terms to see why these questions matter.

As a field matures, new research has to explain what it adds. Its focus may narrow: a particular institution, a certain type of firm, a specific period, or a relationship between two variables that has not yet been adequately explained.

This accumulation can be immensely valuable. Large questions often become tractable only through a succession of smaller ones.

But I think today’s assessment systems too readily treat novelty as a substitute for importance. “Nobody has studied this yet” and “this deserves our time” are fundamentally different judgements.

A young researcher faces publication requirements, citations, job prospects, funding decisions and peer review. The question of why a problem matters can easily find itself competing with more immediate considerations: can the data be obtained? Is there a credible strategy for identifying causal effects? Is the contribution clearly new? Can this become a paper within the time available?

“Why do financial crises recur?” is an important question, but it involves complex mechanisms, limited historical cases and tangled causal relationships. Years of work may not yield a clear answer. The effect of a particular policy on the financing behaviour of a certain type of firm may be easier to define, better suited to available data, and more amenable to a testable research design.

My criticism is therefore not directed at researchers who choose the latter. They may be responding quite rationally to the system. What deserves scrutiny is where that system channels their intelligence.

But if a system consistently favours routes that produce deliverable results, who will take on the work that is important, slow and impossible to guarantee in advance?

9. Universities should make room for long, uncertain exploration

This is one of the tensions in modern academia that troubles me most.

Someone can spend more than a decade mastering sophisticated mathematics, statistics, programming and research methods, only to find that they can deploy those abilities on a very narrow, relatively safe question. The tools advance while the scope of the questions keeps narrowing. I find that a dispiriting waste of talent.

Of course, a question’s significance cannot be measured by how grand its title sounds, or how many people care about it. A seemingly tiny discovery may be a precondition for a much larger breakthrough.

Nor do I think research must make money immediately. Basic research should have room for extensive exploration whose uses are not yet apparent.

The question worth asking is this: do we have institutions that allow capable people to spend ten years on an important problem with a substantial chance of failure?

If researchers must continuously deliver safe, publishable work with clear results, exploration can become subordinate to what is deliverable in the short term.

If universities cannot protect this kind of work, what exactly are they protecting? That is a question I think they owe us a serious answer to.

10. What remains difficult to replace

That is why I do not expect universities to disappear altogether.

In fields such as theoretical physics, materials science and the life sciences, frontier research involves far more than reading textbooks. It requires teams, mentors and accumulated experience. Some of it also depends on extraordinarily expensive instruments and experimental facilities.

It relies, too, on tacit knowledge that is difficult to put fully into writing: how an instrument should actually be adjusted, why a dataset looks wrong, why an elegant idea makes an experienced researcher sceptical, and which apparently promising paths have already been tried repeatedly.

Such judgement develops through sustained practice and is often passed on through working together.

As explaining knowledge becomes easier, the ability to bring people together, maintain research facilities, pass on experience and support long-term exploration may stand out more clearly.

The university’s centre of gravity might move further towards bringing the right people, resources that cannot readily be dispersed, and enough time together to address problems no individual could solve alone.

This would not invent an entirely new mission for universities. It would make some of their most valuable existing functions more important.

11. When knowledge is no longer scarce, what will be?

My judgement is clear: schools can survive as institutions, but their present form, built around standardised instruction, will be replaced.

Schools will lose their position as the default centre of education, along with their authority to define how a person ought to learn.

Today, instruction, social life, credentials, research, selection and childcare are bundled into one institution. In the future, these functions could be recombined: AI could support personalised learning; local communities could sustain shared life; projects and practice could develop ability; universities and research institutions could continue to support inquiry that demands accumulated expertise and close collaboration.

Evidence of ability will also become more specific. Projects, portfolios and records of practical work should carry more of the judgement we currently entrust to a degree certificate.

The more significant change would be that learning no longer follows a single route laid out in advance.

A person could learn something, begin to act, encounter a problem in practice, and then seek the people best suited to solving it together. Those people need not be the same age, hold the same qualifications or belong to the same discipline.

They would come together around something they wanted to accomplish, discuss it, collaborate, fail and continue learning.

When knowledge becomes easier to obtain, the ability to understand, assess and use it does not automatically become abundant. Judgement, creativity, practical experience, the capacity to connect different kinds of knowledge, trust between people, and the ability to recognise a problem worth solving will become the truly scarce resources.

Easier access to answers does not mean worthwhile questions will arise on their own.

Perhaps then we will see more clearly that education should do more than hand someone the answers or train them for a particular specialism.

It should also give them the ability to understand the world, the courage to enter unfamiliar territory, a sense of which problems deserve their commitment, and the chance to find people unlike themselves who are willing to work on those problems together.