AI Can Make You Productive or Dumb. Here’s How to Tell Which Side You’re On

It all depends on how you use AI.

Image made with GPT 4o

Generative AI tools like ChatGPT have seen a massive rise in popularity, with over half of Americans reporting they had tried AI assistants or coding copilots by early 2024. From writing emails to generating code, more and more of our mental tasks are being outsourced to machines.

This raises an important question: why bother remembering facts or composing text when AI can handle it for us? Is this digital convenience making us intellectually lazy—or even less sharp?

Recent studies, expert opinions, and real-world examples suggest that while AI can significantly boost productivity, it may also weaken our critical thinking, memory, and social skills if we’re not careful. The impact isn’t entirely positive or negative; it largely depends on how we choose to use these tools.

Automation and Atrophy: When AI Thinks for Us

One of the clearest indicators of AI's impact comes from a study by the MIT Media Lab, which scanned participants' brain activity while they wrote essays with and without AI assistance. The results were striking. Participants who used ChatGPT showed the lowest levels of brain engagement and consistently underperformed in terms of neural activity, language use, and behavior, according to the researchers.

In short, their brains were less active, especially in areas related to memory, imagination, and critical thinking, compared to those who wrote their essays without AI support.

The essays produced with AI also tended to be more generic and lacked originality. Many included repetitive, formulaic phrases. As participants continued writing, their effort declined. By the third essay, several were allowing ChatGPT to handle nearly all of the writing, simply copying and pasting the output.

The more they leaned on AI, the more passive their approach became.

This mental disengagement came with real consequences. After the writing tasks, participants were asked to recall details from their own essays. Only 20% of those who had used ChatGPT could remember a single quote from their work, compared to 85% of those who wrote without AI or only used tools like Google.

Even more surprising, 16% of the AI users didn’t recognize their own essays when presented with them. None of the participants who had written without AI had this issue.

In effect, AI had disrupted their memory of the task. They completed the assignment, but the learning didn’t register.

“The task was executed efficiently and comfortably,” said Dr. Nataliya Kosmyna, the study’s lead author, “but essentially nothing was integrated into the memory networks.”

When AI handles most of the cognitive load, our brains tend to slow down. It’s a strong example of cognitive offloading and appears to weaken the very abilities we rely on AI to perform.

A separate study by Microsoft and Carnegie Mellon University also found that people solved data-related problems faster with AI tools, but at the cost of critical thinking. Workers with access to generative AI were more efficient, but they showed less reasoning and analysis than those who worked without it.

The study also found a dose-dependent pattern. The more participants relied on AI-generated answers, the more their own critical thinking declined.

Real-world experiences support these findings. Software engineer Koen Van Belle noticed it himself after frequent use of GitHub Copilot. When his internet connection failed one day, he struggled to recall programming syntax and fix problems on his own. “I couldn’t remember how it worked,” he admitted.

He realized he had become overly dependent on AI and had to retrain himself in skills he once knew by heart.

Later, when his company restricted AI access, productivity and code quality among interns dropped significantly. It became clear they hadn’t developed the core problem-solving skills needed to work independently.

This points to an important insight. By skipping the challenge of learning and letting AI generate solutions, many young programmers missed the chance to build the mental resilience needed to solve new problems on their own.

The New “Google Effect”: Memory and Learning in the Age of AI

Relying on AI to do our thinking doesn't just pose a risk to creativity and critical reasoning. It may also quietly reshape how we learn and what we actually retain.

Cognitive science has shown for decades that learning is an active process. We absorb knowledge through effort—by working with ideas, making mistakes, and gradually refining our understanding.

So what happens when an always-available AI steps in and interrupts that process?

Educators are already raising red flags. John Warner, a writing instructor, argues that using ChatGPT to write essays or complete assignments can “derail the important exploration of an idea that happens through the act of writing.”

In other words, when a student skips straight to a polished answer, they miss out on the deeper, more meaningful learning that comes from wrestling with the problem.

This isn’t just a theoretical concern. There’s real evidence it’s already taking place.

A 2023 classroom study found that students who relied heavily on AI to answer questions ended up learning less overall. High school students who used AI more frequently also showed less flexibility when adapting to new social situations.

In an experiment at Duke University, students were asked to summarize readings and then answer questions about them. Those who used AI to generate their summaries performed worse on exams than students who took notes and recalled the material from memory. Researcher Qirui Ju explained that writing things out by hand helps strengthen understanding. When AI takes over that process, you might get a grammatically clean result, but you lose the part where actual learning happens.

As he put it, even if AI-generated responses are more polished, the learning is weaker because students aren’t making the mental effort themselves.

This pattern is really just a more advanced version of what researchers have called the “Google Effect.” We've already seen how internet searches made it easier to forget basic facts. Now, AI doesn’t just help us find answers—it creates them. It can write essays, solve problems, and even generate full solutions, which means we might not even try to come up with ideas on our own.

A study from 2011 found that people often turn to the internet to answer tough questions and, as a result, remember less information directly. What they do remember is how to find the answer. But with AI, even the act of searching and organizing information is being done for us, which can distance us even further from the material we’re trying to learn.

According to a Harvard study titled The Google Effect, the internet has become a “primary form of external or transactive memory.”

In a way, it has become an external brain. AI could take this concept even further, acting not only as a memory tool but as an external problem solver and creative engine. Many argue that it replaces a key part of our thinking process—especially the early, messy stages of idea development that used to happen in our own heads.

If we're not paying attention, we could end up with answers that don’t come from real understanding. In other words, solutions without skills, and knowledge that doesn’t actually belong to us.

A Double-Edged Sword: Using AI Wisely vs. Letting It Replace Us

Despite some of the more cautionary findings, it's important to remember that AI isn’t inherently harmful or designed to dull our minds. That perception still lingers for some, but at its core, AI is simply a tool. The outcome depends on how we choose to use it.

In fact, some research offers hope that AI can serve as a cognitive enhancer rather than a barrier—if applied in the right way.

In the MIT essay study, there was a second phase where the groups switched roles. Students who had written without AI were then allowed to use ChatGPT, while those who had used AI were asked to write on their own.

Interestingly, the students who started by writing with their own knowledge and then used AI later showed strong brain activity and even better performance. Because they had done the mental groundwork first, they were able to use AI more effectively—as a creative assistant or editor—without losing engagement in the process. In these cases, AI seemed to support the learning experience rather than replace it.

The students who began with AI, however, had a harder time when it came time to write without it. Since the AI had done most of the thinking for them at the start, they hadn’t developed any new understanding or creative momentum to carry forward.

This suggests that AI has the potential to enhance human thinking, but only when it's used to complement our own efforts. If we activate our minds first and then use AI to refine, expand, or explore ideas further, we can gain both efficiency and depth. But when we rely on AI from the start, we risk bypassing the very process where real learning happens.

There are also situations where delegating to AI makes sense and doesn’t pose a major risk. A good example is the calculator. Once someone understands basic arithmetic, using a calculator for more advanced math doesn’t make them less capable—it just makes them more efficient. In fact, studies show that there’s little difference in math performance between students who use calculators and those who don’t, as long as the core concepts are already well understood.

The important thing is that the foundation comes first.

Similarly, software engineer Koen Van Belle decided it made sense to let AI write routine social media posts for his company so he could focus on more complex, strategic work. In cases like this, using AI to save time on repetitive, low-stakes tasks can be a smart move. The problem arises when we start outsourcing the core skills that define our work and thinking.

This same tension is playing out across businesses and society at large. Some companies are exploring ways to replace junior developers or entry-level writers with AI to increase output and cut costs.

But the MIT research on programming suggests this might backfire over time. If younger employees never get the chance to build their own problem-solving abilities, creativity and long-term talent development could suffer.

It’s the classic trade-off between short-term productivity and long-term competence.

In education, the same concern applies. Relying too heavily on AI may help students turn in polished assignments today, but it may also produce weaker thinkers tomorrow. Policymakers and educators are currently debating how to draw the line. Some have even raised concerns about proposals like "GPT for preschool," which researchers strongly oppose, warning that short-term convenience should never come at the cost of long-term brain development.

In the end, AI truly is a double-edged sword. It can increase productivity, but it can also reduce motivation to do the work on our own. AI makes it easier to get results, but that same ease can quietly weaken our drive and curiosity.