AI can't change what truly matters in education - despite the hype

In a recent X post, A16z's Andrew Chen asked people to share their thoughts on the future of education with AI.

what happens to traditional education in a world of AI? please speculate.And yes we're going to call it vibeschooling.

— andrew chen (@andrewchen) March 13, 2025

I came across this post just a week after hearing my wife work through the same exercise for an AI education class she took at Stanford, where students were asked to write a short essay sharing their thoughts on the future of traditional education with AI.

In the brief moment between reading Chen's post and clicking on the comments, my mind drifted back to that conversation, and I had a sense of where the comments would go. Sure enough, they went as I'd predicted—many comments revolved, in one way or another, around this common theme of streamlining information by making it faster, more personalized, more engaging or more fun.

Sure enough, I can see the future of learning in Danny's comment, but what I fail to see is how that will fundamentally change the nature of learning - just as all AI's ancestors have failed to do. The fact that the traditional classroom setup has remained largely unchanged since the 1800s, despite significant technological advancements, speaks volumes.

But before presenting my argument, I feel it's important to categorize learning into two distinct groups based on the extent to which knowledge transfer is the primary—or sole—reason people engage in learning: K-12 and post-secondary education and beyond. If you have children, your already know that the closer students are to kindergarten, the more you expect from the school to provide activities way beyond formal learning. Mr. Chen understands this well:

For those obvious reasons, I’m not building my argument based on K-12. Instead, I believe "post-secondary education and beyond" provides an environment with fewer external influences, allowing me to focus on the core of learning. At this level, while socialization and other factors still matter, the primary goal shifts more clearly to knowledge transfer in its purest form.

Many entrepeneurs thinking about AI in the education remain fixated on reimagining content delivery, overlooking the most important factor: human behavior. Enhancing content can only marginally improve learning outcomes—even if those enhancements are dramatically better, which I believe it's the case with AI. A focus on content might disrupt the teaching side of the equation, but not the learning; and this barely relates to K-12 teachers, who as we all know, do far more than just deliver content.

By the time students reach college, they either understand that true learning is driven by human behavior—ideally through intrinsic motivation—or they’ll inevitably encounter frustration.

There's no technological salvation to learning

Let’s be honest—when was the lack of quality resources a bottleneck for learning in the past decade? With world-class, freely available materials just a click away, it's hard to believe that people didn't take time to learn because they didn’t have the resources.

I'm not trying to minimize the impacts of AI—just striving to frame its impact on learning at the right scale. AI has already changed the way I code and my latest experience was vibecoding an entire app using Claude. I haven’t been this impressed by technology since the creation of the Internet, and I'm aware we're still in its early days. But if I can't answer the question of how the current AI solutions are fundamentally boosting learning motivation, I can't see how AI is radically changing learning.

Sure, an AI tutor can guide me through any topic, offering depth and breadth beyond any human instructor. It makes navigating information easier than ever. But the real work—turning raw information into useful knowledge—still falls solely on the learner, and that's the component to learning that matters most.

It’s easy to see how vibecoding, vibewriting, and even vibeschooling could thrive with today’s AI. But vibelearning? That’s a whole different challenge. When you vibecode or vibewrite, you produce actual tangible code or writing as output. But I fail to see a counterpart for learning on those exact terms.

Note to self: change your opinion once "vibelearning" becomes prompting an AI and selecting output you want directly injected into your brain.

Unlike skills that AI can directly augment, learning is constrained by the biological limitations of the human brain... and we haven’t even touched the subject of knowledge retention yet.

I just recently took a Statistics course at Stanford and made a deliberate effort to reinforce each concept as I progressed. My personal study approach involves not moving on to the next topic until I feel the current one has fully sunk in. It’s a mix of strategies—reading, doing exercises, and actively engaging my brain by asking myself questions out loud. This last step is crucial, because it signals to my brain that the information is important and should be stored. For this to be effective, I need maximum focus and avoid distractions at all costs.

While I frequently used ChatGPT for this class, I found myself relying more on traditional study methods—reading textbooks, watching tutorials, and working through exercises. Interacting with AI felt somewhat superficial or distracting, like the doomscrolling equivalent of learning, since it didn’t force me to deeply engage with the material or sift through content thoughtfully. When overwhelmed with information, it's easy to confuse the feeling of learning with actual knowledge retention. Because of this, I had to be intentional about how I incorporated ChatGPT into my study routine to make it a meaningful part of my learning process.

I realize this is more my own challenge than an issue with AI, but I'm not alone in this realization. In fact, experts far more qualified than I am, like Cal Newport, advocate for digital minimalism as a means to enhance learning and focus. See: Deep Work and Digital Minimalism. I can only imagine that for someone with ADHD, AI tutoring might feel overwhelmingly chaotic.

That said, AI tutoring provided me with new ways of exploring content—a honorable mention to item 4, which has provided me with something that wasn’t even possible before:

1. Solving exercises I couldn’t solve on my own.
2. Generating additional exercises when the textbook didn’t have enough.
3. Diving deeper into or expanding on topics not covered in the course.
4. Tailoring explanations to my level of understanding with prompts like “Explain <topic> as if I'm five” or “Explain <topic> as if I'm a high school student.”

I discovered that AI tutoring fits into my study routine in two different ways: at the very beginning, to help plan my study sessions, and at the end, when my focus was too drained for more intensive tasks like reading a textbook or solving exercises.

Despite these clear benefits my overall learning experience resembled more of a traditional way of studying rather than being a complete departure from it. In this sense, AI didn't remove the most arduous process of learning, in fact it barely changed it.

Conclusion

The true bottleneck to learning isn’t access to knowledge; it’s the limited processing power of the biological brain coupled with human behavior. AI processes information like a beam of light racing through the air, but as soon as it hits the water (your brain), it slows down and bends.

While tools can assist, they can't currently replace the core elements of human learning like effort, focus, discipline and long-term dedication. In fact, if used recklessly, AI is more likely to weaken these essential forces rather than strengthen them.

A clear example of this is our reliance on GPS—many people have lost their natural sense of direction to the point where a GPS failure could leave them unable to navigate even routine tasks, like picking up their children from school. If a GPS blackout ever occurs, many would struggle with basic everyday navigation. While I'm fine outsourcing this ability altogether, I refuse to surrender my cognitive faculties in the same way.

Perhaps in the distant future—on the verge of the Singularity—learning, the mother of all skills, for better or worse, will no longer be necessary, at least not as we know it today. What comes next? I have no idea, but I imagine it's around this time that we'll finally witness a fundamental transformation in human learning. Until then, you're better off refining the timeless principles of effective learning.