Is AI Eroding Our Intelligence? Does Generative AI Make Us Smarter or Dumber? — The Inconvenient Reality Revealed by Research

Does AI Really Make Us "Stupid"?

"Does using AI make humans stupid?"

This question might sound a bit harsh. However, in an era where generative AI can write, summarize, propose ideas, organize emails, and even outline debates, it is a question that cannot be ignored.

In the past, we would consult a dictionary if we didn't know something, use a calculator if calculations were tedious, and open a map if we got lost. Now, we ask AI before searching, rely on AI before writing, and say "summarize" to AI before organizing our thoughts.

Is this efficiency, or is it outsourcing our thinking?

An article by German IT journalist Jörg Schieb tackles this issue head-on. The focus is not on the fear of AI itself but rather a warning about the habits of the humans using AI.

AI doesn't make people stupid instantly. However, if the habit of relying on AI before thinking accumulates, the mental "muscles" used for thinking will go unused. Just as muscles atrophy if not used, thinking also dulls if not trained. The problem is not the existence of AI but our continued use of AI as a "thinking substitute."

How is it different from calculators or search engines?

There is a common counterargument from AI advocates.

"When calculators first appeared, people said humans would lose the ability to calculate."
"When search engines emerged, people said memory would decline."
"Every time a new tool comes out, similar anxieties are expressed."

Indeed, there is some truth to this counterargument. Tools have always taken over some human abilities. Paper was a tool to externalize memory. Calculators externalized calculations. Search engines externalized access to knowledge.

However, there is a decisive difference with generative AI.

Calculators primarily take over the limited task of calculation. Search engines take over the task of finding information. Of course, these also impact cognition. However, the final decisions of "which information to trust," "how to organize it," and "what conclusion to draw" still remained with humans.

On the other hand, generative AI handles everything from structuring sentences, organizing arguments, creating counterarguments, assisting in decision-making, to adjusting expressions. In other words, it intrudes into the "thinking process" itself, not just tasks.

This is crucial. AI not only lends a hand but sometimes takes over the very order of thinking.

Answers appear before you struggle. Sentences are completed before you stumble over words. Plausible arguments line up before you search for counterarguments. The more convenient it is, the more people avoid the preliminary struggles.

However, intelligence grows precisely through those struggles.

The Key is "Cognitive Offloading"

To understand this debate, the concept of "cognitive offloading" is important. This refers to entrusting cognitive tasks such as memory, judgment, calculation, and organization to tools outside one's own mind.

Cognitive offloading itself is not evil. Writing a shopping list on a memo, entering schedules into a calendar, saving phone numbers in a smartphone—such actions make daily life easier and help focus on more important matters.

The problem is what we offload.

Not remembering phone numbers might not be a significant intellectual loss. However, if we offload the ability to construct arguments, question others' claims, create the flow of a text, and refine thoughts through trial and error, it's a different story.

The danger of relying on AI is that it easily becomes "outsourcing of thinking" rather than just "outsourcing of memory."

For example, a student might ask AI to write a 2,000-word essay on a topic from the start. An employee might ask AI for good ideas before forming their own hypotheses for a proposal. They might attend meetings without reviewing materials, relying only on AI summaries.

In that moment, it seems efficient. However, the experience of constructing logic, struggling with writing, and selecting information is less likely to remain within the person.

The more AI creates the end product, the more humans become owners of the product but not experiencers of the thought process.

MIT Research Shows Differences in "Brain Usage"

The concern was heightened by MIT Media Lab's study "Your Brain on ChatGPT." In this study, participants were divided into groups that wrote essays independently, used search engines, or used ChatGPT, and their brain activity during essay creation was examined.

The results showed that the connectivity of brain networks was strongest in the group that wrote independently, intermediate in the group that used search engines, and weakest in the group that used ChatGPT. Additionally, those who used ChatGPT had weaker memory and ownership of their written content and struggled to accurately quote their own writing.

Of course, it is premature to conclude from this study alone that "AI makes humans stupid." The number of participants was limited, the paper is a preprint, and further verification is needed. Results may vary depending on the type of AI, the content of the task, how it is used, and the age and proficiency of the users.

Nonetheless, the issue raised by this study is significant.

In the short term, using AI makes writing easier. But how much is the brain involved in that process? How much of it is retained as one's own thought? Even if the task is completed, is learning taking place?

This question relates not only to educational settings but to all intellectual work.

"The More You Trust, the Less You Think"

Microsoft's research also examines the relationship between generative AI and critical thinking. The subjects were knowledge workers who use AI in their jobs.

Interestingly, it wasn't just the amount of AI usage that mattered but also "how much they trust AI." Those with high trust in AI tended to make less effort to think critically themselves. On the other hand, those who trusted their own judgment were more likely to engage in critical thinking even when using AI.

This is a very important implication in practice.

The danger is not in using AI, but in the attitude of accepting AI's output as "probably correct."

AI can be fluently wrong. It mixes facts and assumptions in plausible sentences, makes weak evidence appear strong, and covers logical gaps with beautiful words. That's why users need verification skills.

However, using AI more does not necessarily cultivate verification skills. If the habit of not verifying becomes ingrained, critical thinking weakens.

Opinions on Social Media are Divided

This topic easily generates significant reactions on social media because many people already use AI and simultaneously feel some anxiety about it.

The prominent reactions on social media can be broadly divided into three categories.

The first is voices of heightened crisis awareness.

"If students let AI write their reports, they won't develop the ability to think."
"If you only read AI summaries at work, your ability to read original texts will decline."
"Once you get used to the convenience, you won't be able to construct sentences on your own anymore."

These voices express concerns that AI might reduce human intellectual stamina. In the field of education, there is particular concern about young generations becoming dependent on AI before acquiring the basics of thinking.

The second is a counterargument.

"AI is being made the villain, but search engines and calculators were the same."
"Using tools and losing abilities are separate issues."
"By leaving tasks to AI, humans can focus on more advanced things."

Those who hold this position believe that AI should not be excessively feared. Instead, they see it as an opportunity for humans to focus on creativity and judgment by being freed from simple tasks. In fact, AI reduces work burdens in many situations, such as transcribing meetings, summarizing key points, drafting emails, and creating rough drafts of materials.

The third is the most realistic middle ground.

"It's not about whether to use AI or not, but how to use it."
"You should create the initial idea yourself and have AI provide counterarguments or improvements."
"Instead of making AI a ghostwriter, it's better to use it as a sparring partner."

This reaction is close to the argument of the original article. It is neither a ban on AI nor a praise of AI but a belief that the initiative should remain with humans.

What's interesting in social media discussions is that the more people use AI, the more they focus on "designing how to use it" rather than simple denial. Those who use AI regularly know its convenience. At the same time, they also know the feeling that their own thinking becomes shallow if they leave everything to it.

Therefore, the practical sentiment is settling on "AI is not dangerous" or "AI is not omnipotent." It's becoming "it's dangerous to use without thinking, but powerful if used for thinking."

The Problem is Not AI, but the "First Move"

So, how should we use AI?

The most important thing is not to hand over the first move to AI.

If you're writing, first jot down rough notes yourself. If you're thinking of a plan, first come up with three hypotheses yourself. If you're researching, first define what you want to know in your own words. If you're constructing an argument, first tentatively set your own conclusion.

Then call in AI.

"Point out the weaknesses in this logic."
"List three opposing opinions."
"Highlight points that readers might question."
"How can this structure be made clearer?"
"List the parts that need fact-checking."

With this approach, AI becomes a partner in strengthening thought, not a substitute that takes it away.

On the other hand, if you ask from the start, "write everything," "think everything," "come to a conclusion," AI becomes a convenient substitute. At that moment, work time might be reduced. However, the mental load that should have passed through your own head also disappears.

The disappearance of this load is comfortable in the short term. But if it eliminates the load necessary for learning and growth, it becomes a loss in the long term.

In terms of muscle training, AI can be both an auxiliary tool and an electric wheelchair. If used as an auxiliary tool, it allows you to tackle heavier challenges. However, if you continue to ride it even when you can walk, your leg muscles will weaken.

Younger Generations May Be More Susceptible

Particular attention is needed for generations that are yet to develop their thinking patterns.

Those who have already gained experience in writing and have a foundation in logical structure and critical reading can compare, question, and revise AI outputs. They can relatively easily utilize AI's convenience while retaining their own judgment.

However, what happens if students who do not yet have their own writing style have AI produce a completed text from the start? They might lack the experience of struggling, revising, failing, and reconstructing on their own.

In learning, the important thing is not just obtaining the correct text. Rather, the process of approaching the correct answer and organizing one's thoughts is meaningful.

Submitting a completed text from AI might end the surface-level task. However, little remains within the student. This is a very serious problem for education.

Of course, it's not about excluding AI from education. Instead, future education will need to teach how to use AI.

However, it's not about "how to finish homework quickly with AI." It's about teaching "how to think before asking AI," "how to question AI's answers," and "how to use AI to deepen one's own thoughts."

The Same Thing Happens at Work

This issue is not limited to students; it is also occurring in the business world.

Replying to emails, taking minutes, writing proposals, creating sales materials, drafting research memos, and internal reports—generative AI can make these tasks much easier. For busy workplaces, this is a great help.

However, if everything is entrusted to AI, the abilities that were naturally honed through work may weaken.

For example, writing a report for a superior is tedious, but it includes training in "choosing what is important," "considering the order that conveys the message," "eliminating unnecessary information," and "clarifying conclusions." If you hand it all over to AI, you can skip that training.

The same goes for summarizing meetings. AI summaries are convenient, but if more people rely solely on summaries to make decisions, their ability to read the tone, hesitation, silence, and context of statements might weaken.

Efficiency is necessary. However, attention should also be paid to the learning opportunities lost due to efficiency.

Intelligence in the AI Era is Determined by "How You Ask"

In the future, intelligence will not be measured by the amount of knowledge alone. In an era where AI can quickly provide a lot of knowledge, "what to ask," "how to question," and "how to use" become important.

A person strong in AI is not someone who has AI write everything. It is someone who can judge where the AI output is shallow, where it is dangerous, and where another perspective is needed.

In other words, what is needed in the AI era is not the patience to avoid using AI, but the thinking ability not to be dominated by AI.

To achieve this, the following three habits are effective.

First, write down your hypothesis before asking AI. It can be short. It can be wrong. First, express the thoughts in your head.

Second, ask AI for counterarguments, not answers. Let it not only reinforce your thoughts but also break them. Let it find weaknesses. Let it play another role.

Third, re-explain AI's answer in your own words