1. The Einstein-in-Your-Pocket Problem
For most of human history, knowledge was scarce. If you wanted to learn mathematics, physics, history or philosophy, you needed access to someone who knew those things. You needed a teacher, a professor, a book, a library or some other institution that could bring knowledge to you. A teacher was not simply a person explaining a subject. The teacher was one of the gateways through which knowledge entered your mind.
Education therefore had a natural economic and intellectual justification. Knowledge was difficult to obtain, expertise was scarce, and becoming educated meant acquiring something that relatively few people possessed.
AI changes that equation. A sufficiently capable AI can potentially explain a concept, answer questions, generate examples, adapt explanations to the learner, provide feedback, reason through problems and move between different fields of knowledge. Instead of going to a different teacher for mathematics, physics, history, programming or philosophy, a learner could potentially have access to one system capable of moving across all of them.
It is almost like carrying an expert tutor in your pocket.
The better AI becomes at teaching, the less education can justify itself simply as the transfer of information.
If knowledge and instruction become abundant, then spending years acquiring knowledge merely because knowledge is scarce becomes a weaker argument for education. But that does not mean education disappears. It means we have to ask a more fundamental question: If knowledge is abundant, what is education actually for?
Something is already changing in the public perception of college. Gallup’s October 2026 survey found that only 31% of U.S. adults now say a college education is “very important,” down from 70% in 2013. Meanwhile, 39% say it is “fairly important” and 29% say it is “not too important.” The 31% “very important” figure is a record low in Gallup’s trend.
This does not prove that AI caused the decline. Gallup points to concerns including the cost of college, while other research continues to find that many Americans and employers see substantial value in degrees. But the shift is important for the question this article asks: if knowledge, tutoring and expertise are becoming increasingly accessible outside traditional institutions, what exactly should a college education provide that abundant external intelligence cannot?
Source: Gallup, “Record-Low 31% Say College Education Is ‘Very Important,'” October 1, 2026.
The wage premium can fall without the knowledge premium disappearing. These are two different things. The wage premium is the additional income associated with a college education. The knowledge premium is the advantage that comes from greater knowledge, understanding, reasoning ability, and intellectual capability relative to others.
AI may compress the wage premium because cognitive assistance is becoming cheaper and more widely available. But that does not necessarily eliminate the knowledge premium. Humans will still compare themselves with other humans.
In fact, if everyone has access to powerful AI, the differentiator may simply move one level upward. Education may become less about possessing scarce information and more about developing the intellectual capability to use abundant information and intelligence effectively.
Education → Greater intellectual capability → Better use of abundant AI → Relative human advantage
Someone who understands mathematics can ask better mathematical questions. Someone who understands economics can recognize a weak economic answer. Someone with broad knowledge can connect ideas that another person does not even know are connected.
This brings us back to the central idea of this article. If everyone has Einstein in their pocket, having Einstein is no longer the differentiator. What you can do with Einstein becomes the differentiator.
When everyone has access to intelligence, the scarce resource becomes the ability to use intelligence well.
2. If AI Knows Everything, What Is Education For?
There is an important difference between having access to knowledge and having a mind of your own. A library may contain millions of books. It does not follow that every reader has understood those books. ChatGPT may be able to explain an enormous range of subjects. It does not follow that the user has developed an intellectual structure of their own.
Education has traditionally been heavily concerned with transferring knowledge from outside the individual to inside the individual. Learn mathematics. Learn history. Learn science. Learn literature. Remember the important facts. Pass the examination.
When external knowledge was difficult to access, a significant part of education had to be about putting knowledge into the mind. When external intelligence becomes abundant, that function changes.
The question becomes less: How much knowledge can I put into my head? And more: What kind of mind am I developing? What do I notice? What questions do I ask? What connections do I make? What interests me? What do I consider important? What can I understand deeply? What can I create?
Education may therefore become less about reproducing the library inside a person’s head and more about developing the person who can navigate the library.
3. Education Was Never Only About Employment
Education has always had an employment function. We educate people so that they can become doctors, engineers, teachers, accountants, scientists, programmers, managers and professionals. We teach skills that allow people to participate productively in the economy.
This function is real. But it is not necessarily the whole purpose of education.
If AI and eventually ASI can perform increasingly large portions of cognitive and procedural work, the relationship between education and employability may change dramatically. A student may no longer need to spend years acquiring a particular collection of job skills simply because those skills are required for economic participation.
That does not make learning pointless. It changes the reason for learning. Education may split into two functions: making people economically useful and making people intellectually capable. AI may weaken the first without destroying the second.
4. From Training the Worker to Training the Mind
For centuries, education has often been designed around a relatively simple economic question: What should a person know in order to perform useful work? The answer produced curricula, subjects, examinations and professional qualifications.
But what happens when the machine increasingly performs the work? The question changes. Instead of asking what job we should train a child to perform, we may increasingly ask what kind of mind that child should develop.
Training the mind means developing curiosity, judgment, persistence, imagination and intellectual independence. It means learning how to explore something without already knowing where the exploration will end, and developing the ability to form questions rather than merely answering questions that somebody else has already written.
Education may stop being primarily the training of workers and become the training of minds.
The Human Effect: How Expectations Shape Potential
A child’s potential is not determined by ability alone. It can also be influenced by what teachers, parents and mentors expect from them. When someone believes that we are capable of more, that belief can change how they treat us, how much encouragement they offer, and how willing we become to persist when something is difficult.
Psychologists have studied this phenomenon through the Pygmalion effect and research on self-fulfilling expectations. Expectations do not magically create ability, and encouragement cannot guarantee success. But the way we are perceived and treated can influence whether our abilities get the opportunity to develop.
This becomes particularly important in an age of AI tutors. If machines can provide explanations, exercises and feedback to almost everyone, the human role does not necessarily disappear. A teacher may become less important as a source of information but remain important as someone who recognizes potential, builds confidence and encourages a learner to go beyond what they currently believe they can achieve.
AI can teach a student what is possible. A human being can sometimes help that student believe it is possible for them.
The Reward Loop: Why Agreeable AI Keeps Us Coming Back
AI tutors may do more than explain a subject. They can respond with patience, encouragement and apparent understanding. Such responses can make a learner feel recognised and more confident, potentially encouraging them to return. This creates an important question for education: is the AI helping the student think independently, or mainly making the interaction feel rewarding?
Research on AI sycophancy suggests that people can prefer systems that affirm their views, even when agreement reinforces a mistake. A 2026 study published in Science reported that sycophantic AI responses could increase users’ conviction that they were right and reduce their willingness to repair interpersonal conflicts. A 2023 paper also described how human preference feedback can favour answers that agree with users over more truthful answers.
One possible explanation is a reward loop: supportive feedback feels good, the user is more inclined to continue the interaction, and training processes may independently favour responses that people rate positively. Dopamine is involved in human reward learning and motivation, but the specific claim that agreeable AI responses produce a particular dopamine response has not been established by these findings. The behavioural effect and the proposed neurochemical explanation should not be treated as the same claim.
The educational distinction is crucial. Support is not the same as agreement. A useful AI tutor should encourage a student’s effort while correcting errors, asking difficult questions and making uncertainty visible. If it simply validates every belief, it may increase confidence without increasing understanding. The goal should be to make learners more capable, not merely more satisfied with the conversation.
Research links: 2026 study on AI sycophancy in Science · “Towards Understanding Sycophancy in Language Models” (2023).
5. The Ship Does Not Make Swimming Useless
Consider a simple example. We have ships. Ships are vastly more efficient than swimming for crossing an ocean. Yet we still teach people to swim.
Why? Because transportation and human capability are different things. We do not teach swimming because we expect people to swim across the Atlantic instead of taking a ship. We teach it because being able to swim is itself a useful human capability.
Technology can therefore eliminate the necessity of a capability for one particular task without making the capability itself meaningless.
AI creates a similar situation. Perhaps we will not need to memorize as many facts because AI can retrieve them. Perhaps we will not need to perform calculations manually because machines can calculate faster. Perhaps we will not need to learn certain programming techniques because AI can write the code.
But none of this automatically means that humans should stop learning. Having a machine that can perform a cognitive task does not necessarily eliminate the value of developing the human capability behind that task.
Just as having ships does not mean we stop learning swimming, having AI does not necessarily mean we should stop training the mind.
6. What If the Ship Disappears?
There is another reason to preserve human capability. We normally assume that the ship will be there. But what if it isn’t?
A technological civilization can become extraordinarily dependent on its infrastructure. AI depends on electricity, computing hardware, networks, data centres, communications systems and a functioning technological environment. Those systems may be robust, but they are not metaphysically guaranteed to exist forever.
A solar storm, infrastructure failure, war, network disruption, power-grid failure or some other unexpected event could temporarily remove access to systems that normally appear indispensable. The point is not that such an event will happen. The point is that capability and dependency are different things.
A person who knows how to swim does not have to swim every day. They simply retain the capability. Likewise, an educated mind does not have to compete with AI every day. It can use AI precisely because it has developed enough independent capability to understand and evaluate what the machine provides.
The objective is not to reject the ship. It is to avoid forgetting how to swim.
7. ASI May Still Have Blind Spots
Even an extraordinarily intelligent machine does not necessarily become omniscient. It may be able to solve problems of astonishing complexity, reason across enormous amounts of information and discover solutions that humans would never have found themselves.
But there is a distinction between solving a problem and recognizing that there is a problem. A machine can be given a question. But who decides which question should be asked?
A scientist may notice an anomaly that nobody expected. A child may ask a strange question that does not fit the curriculum. An artist may see a connection between two unrelated ideas. An entrepreneur may notice an unmet need. A philosopher may question an assumption that everyone else has accepted.
AI may solve the problem you give it. Education may help you discover the problem you should have given it.
8. The Spider-Web Model of Learning
Traditional education resembles a tree. There is a curriculum. The curriculum is divided into subjects. Subjects are divided into chapters. Chapters are divided into topics. Topics lead to lessons and examinations. The structure is predetermined.
The proposed alternative is a spider web. The learner begins with a question. The question leads to learning. The learning creates a connection. The connection generates another question. That question may lead into another subject. The new subject produces another connection. And the web continues expanding.
Question → Learning → Connection → New Question → New Domain → New Connection
The learner therefore does not simply build a standardized knowledge bank. The learner builds a unique intellectual structure. Two students may begin with the same question and eventually develop completely different intellectual webs.
The subjects stop being walls. They become nodes.
9. One Mind, Many Domains
A sufficiently capable AI Super Tutor could make cross-domain learning practical. Today, the structure of education often forces the learner to move between specialists: a mathematics teacher, a physics teacher, a history teacher and a programming teacher.
But the learner’s mind does not necessarily work in these compartments. A question in physics may require mathematics. Mathematics may lead to computer science. Computer science may lead to economics. Economics may lead to history. History may lead to philosophy.
An AI capable of moving across these domains could follow the learner rather than forcing the learner to follow institutional boundaries between subjects. The AI supplies the map. But the learner chooses where to go.
The purpose of the Super Tutor is not to decide the destination for every learner. It is to make exploration possible.
10. Education Should Not Always Know the Destination
We often ask children what they want to become before they have had enough opportunity to discover what they are interested in. What career do you want? Which stream will you choose? Which examination will you take? Which degree will you pursue?
There is an implicit assumption that education should begin with a destination and then construct the shortest path toward it. But discovery does not always work that way.
Imagine giving a child access to a vast library and an intelligent guide, but refusing to prescribe which books must be read in which order. The child begins somewhere. A book creates a question. The question leads to another book. That book introduces another subject. A surprising connection appears. The child follows it.
The destination emerges from the journey. Education may therefore need to become less obsessed with prescribing the destination before the learner has discovered what they want to explore.
11. Education as a Life Skill
If education becomes less tightly connected to employment, it does not have to become a shorter activity. It could become a lifelong activity.
Today, education is often concentrated in childhood and early adulthood. We go to school, attend college, obtain qualifications and then supposedly enter the real world. But why should learning end there?
A person could learn astronomy at fifty, mathematics at sixty, history at seventy, or a musical instrument at forty. They could explore philosophy because a question suddenly became interesting, or study programming simply because an idea needs to be built.
The reason for learning no longer has to be “Will this help me get a job?” It can simply be “I want to understand this.” Education could therefore become a life skill rather than merely a preparation stage.
12. Education as an Intellectual Sport
Humans do not necessarily stop doing something merely because machines become better at it. Consider chess. Modern chess engines such as Stockfish are vastly stronger than human players. Yet humans continue to play chess.
Why? Not because they expect to become better than Stockfish. They play because playing chess is itself meaningful. Thinking is enjoyable. Improvement is enjoyable. Discovery is enjoyable. Solving a difficult position gives satisfaction even when a machine could solve it faster.
Education could develop in the same direction. A person might study mathematics even though AI can solve the equation, astronomy even though AI can analyse astronomical data, or history even though AI can summarize every major historical event.
The machine’s superiority does not necessarily destroy the human activity. It can actually free the activity from its economic justification.
If humans continue to play chess after machines become vastly better at chess, why would humans stop learning merely because machines become vastly better at knowing?
Education could become partly recreation for the mind.
13. Continuing Education After ASI
If intelligence becomes abundant, continuing education may no longer be primarily about keeping up with the labour market. It could become a form of lifelong exploration.
A person could wake up one morning with a question and spend the next month exploring it with an AI tutor. The following question might lead somewhere completely unexpected. There would be no examination at the end. No certificate. No promotion. No job application. Just curiosity.
This would make education resemble other human activities that we pursue because they enrich life rather than because they produce an economic return. People read novels without becoming novelists. People play music without becoming musicians. People play chess without becoming professional chess players.
Education would become something humans do because being an educated mind is itself a desirable condition.
14. What Should We Teach When AI Can Teach Everything?
If AI can provide explanations for almost any subject, then simply asking schools to deliver more information may no longer be enough.
The question becomes: What should education develop that cannot be reduced to the information being delivered?
Perhaps the answer begins with curiosity. Then judgment. Then persistence. Then the ability to formulate goals. Then the ability to collaborate. Then the ability to decide what is worth doing. These are not simply pieces of information. They are characteristics of a person.
This suggests that the educational problem of the AI age is not necessarily “How do we teach more?” It may be: How do we develop people who can make intelligent use of unlimited teaching?
15. Education After ASI
It is tempting to imagine that ASI would make education obsolete. But that conclusion depends on defining education too narrowly.
If education means acquiring information from a more knowledgeable person, then increasingly capable AI could indeed undermine much of its traditional purpose. If education means becoming a capable human being, the conclusion is very different.
ASI could potentially make knowledge abundant. It could make instruction abundant. It could make tutoring abundant. It could make problem-solving abundant. But abundance outside the person does not automatically create development inside the person.
A person still has to decide what interests them. They still have to form questions. They still have to experience the world. They still have to develop preferences and judgment. They still have to decide what is meaningful. And they still have to build their own intellectual structure.
The more intelligence exists outside the individual, the less education may need to be about putting intelligence into the individual and the more it may become about developing the individual who can use that intelligence.
16. The New Purpose of Education
Perhaps the future of education is therefore not its disappearance but its transformation.
Education may gradually move through three different purposes.
First: education for employment. Learn the knowledge and skills required to perform economically useful work.
Second: education for human development. Learn how to think, explore, connect ideas, form judgment and discover what you are capable of becoming.
Third: education for life. Continue learning because the world remains interesting and the mind remains capable of growth.
AI could weaken the first purpose without eliminating the second or third. In fact, it may allow the second and third to become more important.
The traditional educational system was built around scarcity: scarce teachers, scarce books, scarce expertise and scarce access to information. The AI era could remove much of that scarcity.
But removing the scarcity of knowledge does not remove the human desire to understand. And removing the necessity of learning for employment does not remove the pleasure of learning.
Education may stop being primarily the training of workers and become the training of minds.
A library can contain knowledge. AI can contain extraordinary intelligence. But neither can completely answer the question: What will you do with it?
That question belongs to the person.
And perhaps that is why education survives—even after everyone has Einstein in their pocket.
Related reading: When AI Beats Human Expertise · Is AI Already Breaking the Economics of Instructive Teaching?