No, AI Doesn’t Mean You Won’t Need to Learn

Note: Rapid Learner—my six-week course that teaches you the best way to learn anything—will reopen on Monday for a new session. If you like my books or challenges and want to master the art of learning yourself, I recommend checking it out.

Perhaps the biggest change in the podcasts I’ve been doing for my latest book, Get Better at Anything, compared to the interviews for my previous book, Ultralearning, has been the focus on AI. Just a few years ago, the topic was rarely brought up, now including a few questions about AI has become almost obligatory.

It’s easy to see why. Generative AI is incredible, and its abilities would seem near-miraculous to machine learning researchers transported here from even just a decade ago.

Yet despite AI’s transformative powers, I’m skeptical that AI will fundamentally alter how we learn and, in turn, the kinds of efforts and strategies we use to learn things well. To be fair, generative AI enables a lot of new tactics, but the fundamentals haven’t changed much.

The Case for (Moderate) AI-Education Skepticism

The safest way to reason about future changes is to look at past changes. While it’s always tempting to say, “this time it’s different,” most of the time, it really isn’t. 

Looking at the numerous fads that have come and gone in educational technology, one thing is striking: Despite the enormous promise of dozens of technologies, the changes to both actual education as practiced in schools and how training and learning operate in the ideal, frankly, have been disappointing.

Video recordings were supposed to obviate the need for teachers. The internet supposedly turned students into digital natives. Computerized tutoring was supposed to accelerate the acquisition of problem-solving skills. Giving out free laptops was supposed to enlighten an entire continent. DuoLingo was supposed to make the iPhone generation multilingual.

I certainly haven’t been immune to the hype. I thought open courseware was going to change higher education. But more than a decade later, I can honestly say it hasn’t changed much at all. Weirdos like me might make use of it, but most people either don’t or can’t.

Given this history, I approach any new educational technology with a dose of skepticism. Of course it’s always possible that things really are different this time, and the way we teach, learn and educate will be forever transformed. 

But I doubt it. 

Why Doesn’t Educational Technology Do More?

You could tell several stories about why educational technology isn’t generally more transformative. Perhaps learning is a red herring, and education is just a signaling contest that isn’t enhanced by greater efficiency. Or it could be that the human element is truly central to learning, and technology that tries to replace it is doomed to fail. 

While those explanations have their merits, I gravitate toward the view that learning isn’t altered so radically by (most) technology, because the true work of learning well hasn’t actually changed much. While technology can tweak processes and change things around the margins, it doesn’t fundamentally alter what you need to do to learn because learning takes place in the brain, and brains haven’t changed all that much.

A person who wants to be a skilled programmer, for instance, can certainly benefit from the fact that ChatGPT can explain code, help with debugging, and guide you to resources. At the margins, I expect it to be somewhat easier for a serious student to learn the skill since there are fewer genuinely “stuck points,” such as spending days hunting down an elusive bug or code failing to compile because the programming environment isn’t set up correctly.

But I don’t expect ChatGPT to fundamentally alter what you need to do mentally to learn to code because learning is still going to involve mastering the key concepts, developing procedural fluency in the language, and searing syntax into your brain. 

Similarly, apps like Duolingo don’t fundamentally alter the work needed to learn a language. You still need to learn all the words and grammar and practice speaking and comprehension until you’re fluent. These tools, at most, can smooth some of the rough edges off of learning. (More typically, however, students will use them to smooth off the necessary roughness of learning and thus never actually learn what they wanted to.)

AI Won’t Reduce the Need to Learn

Just as I don’t believe AI is going to radically change how we learn, I don’t think it will radically alter the need to learn, either.

It’s difficult to make predictions about the most extreme case where we rapidly reach superintelligence, and the machine mind can do everything the human mind can do and more. Perhaps we’ll all live in an AI-generated utopia, or maybe the world will be turned into grey goo.

But in the more reasonable case that AI powers continue to grow, exceeding human beings in some dimensions but not others, we’ll likely be in a collaborative system where both AI and human contributions to useful work are important.

In this case, I see no reason to think AI fundamentally differs from other information technologies. Tools enable human minds to offload some cognitive work while expanding the returns for other kinds of cognitive work. The invention of writing, for instance, reduced the need for verbatim memorization while also vastly increasing the amount of knowledge one could encounter. 

The uncertainty about AI’s eventual cognitive abilities means we benefit from being more flexible about our skills. Since it’s not clear exactly which skills, domains and professions will be most heavily transformed, the smart bet is developing the flexibility to learn well into the future.

The Fundamentals of Learning Well Haven’t Changed

Given the uncertainty of what skills will be most useful in an AI age and the high probability that learning will continue to rely on the same processes in the near future, I think that’s as good an argument as any for spending some time learning how to learn.

The human brain is (still) the most sophisticated learning machine on the planet. Unfortunately, it does not come with an instruction manual.

This lack of instructions can cost us. Many of our intuitions about learning are wrong. We think we’ll retain more if we re-read our notes, even though practicing recall is better. We feel like repetitive practice is more efficient than mixing things up (it isn’t). And we continue to cram, even though that’s the surest way to forget everything right after the test.

Worse, we’re terrible self-directed learners. While many of us manage to piece together a decent strategy to pass a test after years of formal schooling, most of us are at a loss when we need to learn new skills after graduation. The art of designing effective learning projects, gathering materials, choosing strategies—and actually sticking with it!—is woefully undeveloped.

A New Session of My Course, Rapid Learner

As many of you know, I teach a course, Rapid Learner, that aims to fill these gaps. I have been teaching this course for nearly a decade. On Monday, I’m opening the course for a new session.

The course is different from my books. While my focus in my published books is to try to introduce useful ideas about learning to a broader audience, there’s a lot I can do in courses that is hard or impossible to do with a published book. That includes pacing the course over six weeks—so you have time to turn it into action, using interactive worksheets—so you can apply the advice step by step, and having the opportunity to interact with students via comments and replies—so you can ask questions and get feedback.

If you’re ready to build the timeless art of effective learning, I hope to see you there!