The AI Paradox: When AI Can Answer Almost Anything, What Is Learning For?
Admissions Enquiry
There was a time when learning began with finding information. To understand something, a student had to search for a book, find the relevant chapter, make notes, ask questions, compare explanations and gradually build an understanding. Access itself involved effort.
That world has changed. AI can explain a concept in seconds, generate examples, compare alternatives, write code, analyse information and produce a polished first draft. Information has never been more accessible. But this abundance raises a different educational question: are we making enough effort to turn information into our own knowledge?
When Information Is No Longer the Challenge
A recent decision in New York City brings this question into focus. For the 2026–27 school year, the city has introduced a one-year moratorium on student-facing generative AI for children from K2 through K8, while allowing more guided use for older students.
The rationale is interesting. The concern is not simply that AI can produce incorrect answers. The policy discussion emphasises the importance of human interaction, curiosity, creativity, independent problem-solving and the experience of working through difficult problems. At the same time, older students are being prepared to use AI responsibly.
The distinction matters because checking AI is not the same as learning. We often tell students to verify AI outputs because AI can make mistakes. That is sensible, but it cannot be the whole argument. AI systems are becoming increasingly capable. Even if they become highly accurate, learners will still need to develop the ability to think, create, solve and make judgements for themselves.
Learning With AI, Not Through AI
Consider something as ordinary as writing an email. A student can ask AI to write it, read it and send it. Or the student can first write a basic version, ask AI to improve it and then compare the two. What did AI express better? What had the student overlooked? Which suggestions fit the intended audience? What would the student retain or change? The second process uses AI, but the learner remains actively involved in learning.
The same principle can work across disciplines. A student can attempt a problem before asking AI for a solution and compare the approaches. A design student can create an initial concept and then explore alternatives with AI. A management student can formulate a recommendation and ask AI to challenge it. A student learning to code can first reason through what a programme should do before asking AI to generate it. In each case, AI becomes a peer to learn with, rather than a substitute for the thinking that creates the learning.
The Skills We Still Need in an AI-Powered World
There is an interesting parallel in the world of AI itself. Forward Deployed Engineers, or FDEs, are an emerging and highly compensated role at companies such as OpenAI. They work closely with customers to identify useful applications, understand workflows, build and deploy solutions and move them into actual use. Their success is not simply whether a prototype works; it is whether the technology creates measurable value in the customer's environment.
Why is such a role needed when the underlying AI is already so powerful? Because there is often a gap between what a model can do and what a real situation requires. The requirement may be unclear, a workflow may not fit neatly, or a new problem may emerge that was not anticipated. FDEs need technical expertise, but also problem-solving and judgement to identify these grey areas and make the technology work in context.
That offers an important lesson for education. If learners routinely outsource the first attempt, the formulation of a problem or the effort to construct an explanation, they may get fewer opportunities to practise the very abilities needed when technology reaches its limits. These abilities are not valuable only for catching AI errors. They help us recognise a new problem, ask a better question, anticipate an unfamiliar situation and create something that did not exist before.
Keeping the Human in the Learning Loop
This is why deliberate learning becomes more important, not less, in an AI-rich world. We can use AI to accelerate exploration, provide alternative perspectives and improve our work. But we also need to deliberately pause, attempt, compare, reflect, explain and revise. Learning should sometimes begin without AI precisely so that AI can subsequently become a source of comparison and discovery: What did I miss? What did I understand differently? What can I learn from this comparison?
Creativity belongs in this conversation too. Creativity needs room for experimentation, imperfect first attempts and unexpected connections. If every first attempt is outsourced, we may gain efficiency while losing some of the space in which ideas are formed. The goal is not to reject AI, but to ensure that convenience does not replace curiosity.
For a long time, the scarcity of information forced us to make an effort to find and understand it. AI has removed much of that friction. Our challenge now is to make a conscious effort to assimilate what we encounter.
The goal is therefore not merely to preserve our ability to validate AI outputs. It is to preserve our capacity to think, create, solve and learn- and then use AI as a powerful peer that extends those capabilities. As AI becomes better at doing the work, we have to take greater ownership of learning how to do it ourselves.