Learning in the Age of AI
Ten Questions about Learning and my First Principles thinking on possible answers to those.
I like reading physical books not a Kindle, nor PDFs on a laptop screen. And I read slowly, deliberately: often I'll stop on a paragraph, sometimes a single sentence, and go back over it before moving on. Not because I missed something, but because I want to sit with it. That habit makes me a little suspicious when I hear people boast about finishing a book a week. It's possible, sure. But speed and rereading are opposites, you can do one or the other, not both and I'd guess a good deal of what makes an idea actually yours happens in the rereading, not the first pass.
For fiction, that may not matter much, as long as the characters, events, and shape of the story stay with you. But for nonfiction, history, economics, psychology, finance, the habit of doubling back is often where the real value sits. There are ideas buried mid-paragraph, in a chapter you were only ever going to skim because your reading app told you to finish the book in nine days. Reading against a clock and reading to understand are two different activities, and we've quietly started treating them as the same one.
That habit of mine; stopping, rereading, contemplating is what sent me thinking about learning itself. Not just how we read, but how we'll learn at all, now that AI can read everything, instantly, for us. If rereading a sentence is where understanding actually gets built, what happens to that process once a tutor is always available to just tell you the answer the first time you ask?
The premise is simple enough to state plainly: the frontier models have already absorbed more of the written world than any human could in a hundred lifetimes, and they keep getting better at reasoning, arguing, and defending ideas in many cases, better than we do. Given that, I put ten broad questions about learning and teaching to a number of teachers and professors I know. None of them have answered yet. So here is my own attempt, working from first principles, rather than wait any longer.
1. What is Learning now and what will it become?
The early stages won't change much. Children will still learn language and basic skills at home, then in play school, kindergarten, and primary school alongside a steady diet of rhymes, comics, and videos on YouTube and similar platforms. I've called this the 21st-century nanny: in nuclear families where both parents work, screens fill the caregiving gaps that grandparents once filled.
What changes is what comes after. Once children reach middle and high school, they'll carry personal AI tutors available 24/7 alongside their tablets. Classroom teaching will lean further into video, graphics, and activity. Textbooks will still exist, but more as a manual connecting what happens in the classroom to what the AI tutor reinforces at home, followed by a quiz that feeds into continuous assessment.
Reading and writing, as standalone disciplined habits, may become one of the more sparsely used skills of the next generation.
2. Is struggle necessary in Learning?
I think so. The discomfort of reading something, failing to understand it, and reading it again is the same habit I described at the start of this piece and is not incidental to learning. It is learning, or at least a large part of the mechanism by which understanding sticks. An AI tutor could, in principle, be designed to preserve that friction. Whether it will be is a different question, and I'm doubtful if friction is exactly what most consumer AI products are built to remove, because removing it is what makes them feel good to use.
If a standardized AI tutor smooths over that struggle for the sake of a better user experience, it risks producing something that looks like understanding but isn't a surface fluency that collapses the moment a later concept depends on the one that was never actually learned. Knowledge retrieved at the click of a button is not the same as knowledge earned.
3. Are we losing deep reading and writing?
Not yet, not entirely; reading and writing still anchor most assessment. But the trend line is visible. Professionally, writing by hand has all but disappeared; we compose on keyboards and touchscreens now, and increasingly, AI drafts the first pass for us.
Deep reading demands sustained, voluntary attention, and that habit is mostly built by reading books in a domain over years. Large parts of Gen Z and even Millennials haven't built that habit; their attention lives on social media instead. I don't think deep reading disappears entirely, but it becomes a minority skill, something a smaller, self-selected group actively cultivates rather than something the system produces by default.
4. What should students still learn for themselves?
Beyond formal coursework, the experiential layer, projects, internships, part-time jobs, workshops, non-academic pursuits matters more than ever, precisely because AI is compressing the value of pure content mastery. What it can't easily hand a student is judgment earned through doing something and getting it wrong.
Language learning belongs in the same category. Fluency in more than one language domestic and foreign, remains one of the more durable investments a student can make in themselves, AI translation notwithstanding.
5. How should teachers use AI?
Professor Steve Joordens, a guest on my podcast, built an AI replica of himself that's available around the clock to students studying psychology under him. I expect versions of this, subject-specific AI instructors modelled on real faculty to spread across schools and universities.
He's also argued, in several conversations including ours, for retiring the multiple-choice question in favour of AI-enabled oral assessment. It sounds ahead of its time today. I think the direction is right, even if the timeline is optimistic.
6. How should pedagogy change?
I recently read a piece in The Hindu by a professor Dr. Aruna Parandhama of English at St. Joseph's College of Commerce, Bangalore, who used a phrase that stuck with me: teach less, steer more. As AI absorbs more of the content-delivery function, the teacher's role shifts toward facilitation, steering a room's discussion rather than transmitting the material for the first time.
That implies a flipped structure: students engage the AI tutor the day before to build a working understanding of the concept, and the classroom becomes the place where that understanding gets tested, through argument, counter-argument, concession, and eventually shared understanding.
7. What happens to assessment?
Joordens again: a future without multiple-choice questions, replaced by continuous assessment woven into each student's AI-tutor relationship. I'd expect mid-terms, end-terms, and viva voce to persist, but the underlying target of assessment to shift, from recalling a date or a formula to demonstrating that a concept can be applied.
Salman Khan of Khan Academy has floated something more radical: a mastery standard where anything below 100% signals an unresolved gap, not an acceptable score. It's a genuinely different way of thinking about what a grade means, and I think it's worth taking seriously rather than dismissing as unrealistic.
8. Will AI widen educational inequality?
In the short run, almost certainly yes. Access to frontier-model tutoring isn't evenly distributed, and the students who have it will pull ahead of those who don't. I don't think a fully equal global rollout of AI-in-education is realistic, that's a utopian target, not a plan. What's more plausible is a slow narrowing of the gap as internet access, device penetration, and AI availability keep expanding, the way mobile connectivity did over the last two decades. Slower than we'd like, but not nothing.
9. What becomes uniquely human?
Knowing everything is, and always was, impossible for humans, and now increasingly redundant to compete on, given AI's reach. What doesn't transfer to the machine as easily is the ability to ask the right question, and to reason about a genuinely open problem from first principles rather than from precedent. That, more than any factual repository, is the more durable human edge.
But that edge isn't free-standing, it's built the same way the rereading habit I opened with is built: through friction, repetition, and the discomfort of not immediately understanding something. A good question doesn't arrive out of nowhere; it comes from having sat with a bad answer long enough to notice what's wrong with it. Which means Q9's answer depends on Q2's: if AI tutoring succeeds in engineering away all the struggle, it may also quietly engineer away the only mechanism that produces the "uniquely human" capacity we're counting on to survive it.
10. If you were designing education from scratch today...
I'd start by dismantling the board-exam and entrance-exam machinery that currently defines success. Both have produced a culture of cramming rather than understanding, and I think that culture does more damage than good.
Beyond that:
Respect different paces. Forcing every student through the same learning speed and the same assessment calendar disadvantages a large share of them for no good reason.
Trade classroom hours for real-world exposure. More project work, internships, and part-time jobs; fewer hours in seats.
Ban the coaching industry in its current form, it exists to game exams, not to educate.
Cut the curriculum aggressively. Drop subjects and chapters a student is unlikely to ever use professionally.
Anchor the foundation in critical thinking and languages. Alongside basic math and science, build in psychology, economics, and social science as core not electives through school. Specialization can wait until after.
These are first-principles answers, not settled ones, I'm still waiting to hear what the teachers and professors I originally asked have to say. If you disagree, don't skim past it: reread it, sit with it, and tell me where it's wrong. I'd rather be argued with than agreed with.
About the Author:
Dr SP Mishra is the Founder of India Career Centre, a Hyderabad-based career guidance, study-abroad consulting and educational research organisation. He writes and speaks on career decision-making, higher education, skills, employability and the future of work in India. Also hosts the India Career Centre Podcast. Connect on LinkedIn.
Comments