As schools rush to respond to artificial intelligence, much of the discussion has understandably concentrated on immediate practical questions: whether pupils should be permitted to use generative AI for homework, how teachers can distinguish a pupil's own work from machine-generated text, whether assessment needs to be redesigned, and whether AI can reduce workloads or provide personalised tuition. These are real and pressing questions, and schools cannot afford to ignore them. But beneath them lies a more fundamental question that is receiving far less attention: what is learning for, and what does it mean for a human being to be taught by a machine?

Generative AI makes this question urgent because it exposes an ambiguity that has quietly run through modern education for a long time. On one view, education is essentially the transmission and demonstration of information: pupils acquire facts, procedures and correct answers, and their progress is measured by how reliably they can reproduce them. On a different, older view, education is the formation of a person: the gradual cultivation of judgement, character and the capacity to think for oneself, of which correct answers are only ever partial evidence. Most schools have always drawn on both views without fully reconciling them. AI forces the question because it is devastatingly good at the first kind of education and almost entirely useless at the second: it can summarise a difficult text, explain a mathematical procedure, generate an essay plan, correct grammar, translate languages and produce apparently personalised explanations, all with remarkable fluency. Faced with this, it is tempting to treat AI simply as a more efficient version of what schools were already doing.

But learning has never been only the acquisition and manipulation of information. A pupil who receives the right answer has not necessarily understood why it is right; one who produces a fluent essay has not necessarily developed the capacity to think; and one with instant access to enormous quantities of information has not thereby acquired knowledge, still less wisdom. Genuine learning involves attention, effort, questioning, judgement, memory, imagination and the slow formation of intellectual habits. It sometimes requires struggling with a problem rather than immediately obtaining its solution—and the educational value may lie precisely in the struggle that technology now allows us to bypass.

This creates a real risk for schools, not a mere inconvenience. A technology built to assist learning can, used without educational judgement, remove some of the very processes through which learning happens. If a machine continually summarises what pupils should read, supplies the arguments they should construct, or produces the sentences they should struggle to express themselves, it may raise the quality of the immediate product while quietly weakening the person who produces it. The question is therefore not simply whether AI improves performance, but what kinds of human capacity its use strengthens, and what kinds it allows to atrophy.

It might be objected that this concern is not new. Writing itself was once accused of weakening memory; the calculator was once feared as the end of mental arithmetic; search engines were said to be destroying the discipline of research. Education has absorbed each of these technologies without collapsing, and AI, on this view, is simply the next tool to be assimilated. There is something to this. But earlier technologies mostly automated storage and retrieval — the memory function of the mind — while leaving the work of judgement, argument and expression to the learner. Generative AI is different in kind, not only in degree: it does not merely store what a pupil might otherwise memorise, but performs the reasoning, drafting and evaluating that were previously the exercises through which judgement itself was built. That is what makes the present moment genuinely, and not merely rhetorically, different.

There is a further issue, beyond what is learnt: education is also about the relationship through which learning takes place. To be taught by another human being is to encounter someone who does more than transmit information. A teacher notices hesitation, recognises potential, challenges complacency, encourages perseverance, and exercises judgement about what a particular pupil needs at a particular moment. The teacher brings personality, experience, moral judgement and human concern into the encounter. Even correction has a relational dimension: a pupil gradually learns that standards matter because another human being takes both the subject and the pupil seriously.

An AI system can simulate aspects of this relationship with extraordinary sophistication. It can respond patiently, adapt its explanations, offer encouragement, and converse in language that sounds sympathetic and attentive. But simulation is not relationship. The machine does not care whether the pupil understands, flourishes or becomes a good human being. It bears no responsibility towards the child whose questions it answers. It has not earned wisdom through living, and it cannot exemplify courage, integrity, compassion or intellectual humility—it can only produce the language associated with these qualities, without possessing them.

This distinction matters because education has always involved formation as well as instruction. Children learn not only from what teachers say but from what teachers are. They encounter adults whose conduct shows them, imperfectly but directly, what it means to listen, to disagree respectfully, to admit error, to exercise authority fairly, to persist through difficulty, and to care about truth. None of this is information that can simply be downloaded. It is cultivated through relationships, practice, example and participation in a community.

The challenge AI presents should not lead schools either to reject the technology or to embrace it uncritically. Used well, it may become a genuinely valuable educational instrument: helping teachers prepare materials more efficiently, giving pupils a patient partner for practising a skill, widening access to explanation for those who cannot get it elsewhere, or supporting a pupil with additional needs in ways a single teacher, stretched across thirty children, sometimes cannot. But the criterion for its use must remain educational rather than technological: does this use of AI contribute to the development of the human being we are trying to educate?

That question requires schools to recover a language of educational purpose broader than efficiency, attainment and employability. Education should certainly equip young people with knowledge and skills, including the ability to use new technologies intelligently. But it must also cultivate discernment: the capacity to distinguish information from knowledge, plausibility from truth, assistance from dependence, and what can be automated from what ought to remain a human responsibility. In an age when machines can generate increasingly convincing answers, knowing how to ask good questions, evaluate claims and exercise independent judgement may matter more than ever, not less.

Perhaps the deepest significance of AI for education, then, lies not in what the technology can do, but in the questions it forces schools to ask about themselves: what capacities are distinctively worth cultivating in a human being, and what do we lose when an encounter that once took place between two persons becomes an interaction between a person and a system?

Schools undoubtedly need policies on artificial intelligence: guidance on assessment, academic integrity, safeguarding, privacy and appropriate classroom use. But policies alone will not be enough. The more capable our machines become, the more clearly we shall need to understand the purposes that should govern their use. The fundamental educational question is therefore no longer simply, "What can AI do for learning?" It is also, and more importantly, "What kind of human being should learning help to form?" Only when we can answer that question will we know what place artificial intelligence should occupy in the education of our children.