Solve 12 Problems With AI, Then Quit Faster Alone

After Solving 12 Problems With AI, People Gave Up Faster on the Next 3 Alone

Socrates worried about writing. He feared that once people started putting things down on paper, they’d stop bothering to remember anything. When the telegraph arrived, there were predictions that poetry would die. When calculators became common, people worried arithmetic skills would vanish. And when Google showed up, the same anxiety played out all over again.

The pattern never changes. Every new tool brings the same fear—“this is going to wreck our brains”—and, for the most part, the disaster never materializes. So when people say “AI is making us stupid” these days, it sounds like the same old script.

But when I actually went through the recent studies one by one, I couldn’t shake the feeling that this time might be different. Generative AI doesn’t just help with looking things up or doing calculations—it helps generate ideas and structure writing. Early research is starting to show that after getting this kind of help, people may struggle when they have to tackle a task on their own.

🧩 The Google and Calculator Debates, Revisited

In 2011, a research team at Columbia University ran a fascinating experiment. When people were told a piece of information along with “you can look this up on a computer later,” they remembered where to find it far better than the information itself. The team called this the “Google Effect.” We ended up remembering how to retrieve information rather than the information itself.

Back then, too, worries and rebuttals ran neck and neck. One camp argued “Google is making us stupid” (a famous 2008 cover story in The Atlantic), while the other countered that “the time we saved from digging through libraries could go toward deeper thinking.” Calculators sparked a similar debate. Our mental arithmetic may have weakened, but calculators let us tackle far more complex mathematics.

That’s why so many people compare AI to a calculator. Sam Altman is one of them. But someone pushes back hard against this analogy: Nataliya Kosmyna of MIT, who produced the most widely cited study on AI and cognitive ability. “You don’t fall asleep hugging a calculator and wake up the next morning. You don’t pour out your inner worries to a calculator, either.” Her point is that the very “personality” of the tool is different.

📊 What Recent Studies Have Found

Let’s start with Nataliya Kosmyna’s team’s experiment. They split 54 participants into three groups. One group wrote essays using a large language model¹ like ChatGPT, another used Google Search, and the last used nothing but their own heads. When they measured brain waves (EEG)², the group that used no tools showed the highest neural connectivity while performing the task, and the LLM group showed the lowest. This measurement alone doesn’t tell us about intelligence or brain damage. But the people who wrote with AI also struggled to accurately quote sentences they had just written, and reported the lowest sense of “this is my own work.” And the gap only widened across sessions.

This isn’t just a lab phenomenon. A Wharton School research team gave AI math tutors to roughly 1,000 high school students in Turkey. One was an ordinary ChatGPT-like tool; the other was a “guardrailed” version that offered hints instead of straight answers and incorporated teacher-written solutions and error logs. During practice, both groups did well—students using the guardrailed version got a striking 127% more practice problems right. But when the AI was taken away and students sat for a test, the picture flipped. Students who had used the plain ChatGPT-style tool actually performed worse than students who hadn’t used AI at all. The practice gains that AI produced didn’t carry over into test performance without the tool.

A recent study by Carnegie Mellon’s Grace Liu and colleagues captured a similar shift on a much shorter timescale. Participants worked on fraction problems; one group could use AI help for the first 12 problems but had to solve the last 3 alone. The AI-assisted group did well on the first 12—but on the final 3, they made more mistakes and gave up more often. This shift took just 10 minutes to appear. The researchers described today’s AI this way: “A mentor doesn’t just hand you the answer. A mentor designs the learning experience and puts your growth ahead of immediate results. Today’s AI, by contrast, is a shortsighted collaborator optimized to give instant, complete answers. It never once says ‘no.’”

It’s a familiar pattern—we’ve already lived through something similar with GPS. A 2020 McGill University study found that the more someone had relied on GPS, the worse their spatial memory was when navigating without it. When researchers measured again three years later, the people who had used GPS more in the interim showed a steeper decline in spatial memory.

What about creativity? A Georgetown University research team compared over 370,000 college application essays from before and after ChatGPT’s arrival. Interestingly, essays with AI involvement used more elaborate and varied vocabulary. But the “ideas” inside them converged toward sameness. The language got richer while the thinking narrowed toward a single point. Adam Green, who led the research, put it this way: “Google helps you find what you were already looking for. AI, on the other hand, decides for you what to look for in the first place.”

I should flag the limits of this body of research. Sample sizes and study designs vary widely—some are short lab experiments, others large-scale observational studies, and some haven’t yet gone through peer review³. Liu herself was careful to draw a line: “Ten minutes of use doesn’t cause long-term brain changes or cognitive decline. What happens with repeated, long-term use is still an open question, and answering it requires longitudinal research.” So it would be premature to declare, based on current data, that “AI is ruining our brains.” Still, it’s hard to dismiss the fact that different teams, using different tasks, keep picking up signals pointing in the same direction.

How AI Differs From Previous Tools

The difference I keep noticing is that AI can take over not just the search but the process of generating and connecting ideas. Writing and search engines shaped how we think, too, but generative AI, when asked a question, hands back an answer already packaged as an argument with supporting reasons. If users accept that structure without examining it, it’s easy to skip the step of thinking it through themselves.

The problem lies with people who haven’t built that capacity in the first place. Michael Gerlich of Swiss Business School warns: “There’s a real risk that younger generations never learn critical thinking from the ground up. The convenience of having AI think for them could mean the skill never develops at all.” Even those who’ve already mastered it aren’t safe. A 1971 study of pilots found that manual, hand-and-eye skills held up fine after a layoff from flying, but the “cognitive” skills — remembering procedures, mentally tracking the aircraft’s position — dulled quickly. The cognitive skills of remembering procedures and tracking the plane’s position eroded before the manual, hands-on skills did.

Google’s precedent is worth chewing on, too. Did Google actually make us smarter or freer? Not really. As digital tools spread, so did worries about blurred work-life boundaries and shrinking attention spans. IQ scores, which had risen roughly 3 points per decade throughout the 20th century, started falling on several measures after 2006 — and the steepest decline showed up among 18–22 year-olds, the very cohort closest to being true digital natives. That said, generational score gaps or changes occurring in the same period aren’t enough on their own to conclude that Google and digital devices were the cause.

There’s one hopeful clue, though. Researchers see shortened attention spans less as evidence of rewired brain structure and more as a matter of “habit.” That implies that if we clear away the distractions we’ve built into our own environment, we can retrain ourselves to concentrate for longer stretches again. If the capacity hasn’t vanished but is merely sitting unused, then there’s just as much room to get it back.

Oswarld’s Lens

Honestly, I’m a little skeptical of both sides on this one.

As someone who’s worked with data, I think the “AI is rotting our brains” headlines are still standing on thin evidence. You can’t declare humanity’s future settled based on experiments with a few dozen subjects, 10-minute observation windows, and papers that haven’t even cleared peer review yet. Anxiety-triggering claims tend to spread faster than the evidence that would justify them. But at the same time, the optimism of “we were fine after calculators, so we’ll be fine now too” strikes me as lazy. If similar problems keep showing up across multiple studies, that’s exactly when we need to dig into what usage patterns and conditions produce those results.

There’s one thing I’ve confirmed over and over again while building AI tools myself: today’s AI is designed to give you an immediate, complete answer. As a product feature, that’s a strength — users get results right away instead of wandering around stuck. But that’s not how learning works. You grow your capacity to think for yourself by wandering, getting stuck, and getting things wrong. As the Liu team put it, AI never says “no.” That’s not a flaw in the AI — it’s a design choice. And design choices can be changed.

So here’s where I land: rather than deciding whether to use AI or not, what matters is deciding for yourself which tasks you’ll keep doing by hand and which ones you’ll hand off to AI. Kosmyna said she uses the AI she built strictly for research, and “proudly” doesn’t touch language models in her personal life. Green said to “respect the blank page.” I try to do the same — at least in the moment of drafting, I switch AI off. I might lose a bit of speed in the output, but I want to keep the sense of thinking itself as mine.

Closing

Here’s the summary. First, the fear that “new technology ruins the brain” has always been around, and it’s mostly missed the mark. Second, difficulties were observed in tasks where people had to solve problems or generate ideas on their own after getting help from AI. We also need to keep in mind that outcomes can differ depending on how the tool is designed and used. Third, it’s still too early to draw firm conclusions—but that’s exactly why it’s worth deciding now what abilities we want to protect as our own.

Why not try this, just for today: pick one problem that’s giving you trouble and stick with it to the end, without AI. Before you check the answer, try explaining to yourself exactly where you got stuck. That alone can be good practice.

💬 Is there a skill you’ve already “handed over” to AI? For me, it’s my sense of direction—I’ve completely outsourced it to GPS. On the flip side, if there’s something you’d say “I will never hand this over to AI,” tell me why in the comments. It might just become material for the next issue.

---

📨 If someone around you has been saying “my head just isn’t what it used to be,” why not pass this piece along to them?

---

References & Further Reading

Primary sources

• Nataliya Kosmyna et al., “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”, MIT Media Lab, 2025.
• Hamsa Bastani et al., “Generative AI without guardrails can harm learning: Evidence from high school mathematics”, PNAS, 2025.
• Grace Liu et al., “AI Assistance Reduces Persistence and Hurts Independent Performance”, 2026 (preprint).
• Louisa Dahmani & Véronique Bohbot, “Habitual use of GPS negatively impacts spatial memory during self-guided navigation”, Scientific Reports, 2020.
• Adam Green et al., “The Link Between Diverse Words and Original Ideas Is Weakening in the AI-Era College Admissions”, PsyArXiv preprint (Georgetown University), 2026.
• Michael Gerlich, “AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking”, Societies, 2025.

Background

• Betsy Sparrow, Jenny Liu & Daniel Wegner, “Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips”, Science, 2011. : This is the original paper behind the “Google Effect.”
• Nicholas Carr, “Is Google Making Us Stupid?”, The Atlantic, 2008. : This piece feels like a deja vu of today’s debate. It’s 17 years old, but it reads exactly the same now.

---

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

📝 Glossary

Footnotes

1. Large Language Model (LLM): An AI, like ChatGPT, that generates text after training on massive amounts of it. It’s the core engine behind what we commonly call “generative AI.” ↩

2. Electroencephalogram (EEG): A method of measuring the brain’s electrical activity using sensors attached to the head. It lets researchers see how actively different regions are “working together.” ↩

3. Peer review: The process by which experts in a field verify a paper’s methods and conclusions before formal publication. A paper that hasn’t gone through this yet is called a “preprint.” ↩