Can AI Actually Make Us Smarter?

Artificial intelligence can accelerate learning, sharpen problem-solving and expand human capability. But the same technology can also encourage us to stop thinking for ourselves. The difference may come down to one question: Are we using AI as a cognitive amplifier—or as a cognitive crutch?

By AI TV INFO | Global Intelligence — Technology & Society

Whether artificial intelligence can actually make us smarter is becoming one of the defining questions of the technological era.

Generative AI can explain quantum physics in seconds, challenge an argument, debug computer code, summarize hundreds of pages and generate a first draft of an essay or business plan. It can function as a tutor, research assistant, editor and intellectual sparring partner—all in the same conversation.

But there is a paradox at the heart of the technology.

The easier it becomes to outsource thinking, the easier it may become to stop practicing it.

The answer, therefore, depends on what we mean by smarter.

Are we talking about what a person can accomplish with AI? What they can remember and understand without it? Or what humans and machines can achieve together?

Those are three different questions—and increasingly, the evidence suggests they have three different answers.

AI as a cognitive amplifier

Used deliberately, AI can resemble an extraordinarily patient tutor.

It can explain the same concept at multiple levels, generate examples, identify gaps in an argument and respond immediately to follow-up questions. Instead of spending hours searching for an explanation that fits, a learner can ask the system to change its approach until the idea becomes clear.

That creates several potential advantages.

Faster learning

AI can break complex subjects into manageable steps, provide analogies and adapt explanations to the questions a learner asks.

Better problem-solving

Rather than simply supplying an answer, AI can be used to challenge assumptions, suggest alternative approaches and identify weaknesses in a proposed solution.

More intellectual exploration

A student, researcher or professional can investigate competing explanations, test hypothetical scenarios and explore unfamiliar subjects without having to navigate every barrier to entry alone.

Faster feedback

AI can critique writing, code, arguments or calculations and explain why something might be improved.

More cognitive capacity

By handling routine tasks—formatting information, debugging syntax, organizing notes or summarizing documents—AI can leave people with more time for strategy, creativity and judgment.

But there is an important distinction.

Increasing what someone can accomplish is not necessarily the same as increasing what they understand.

The performance–learning paradox

Some of the most important research on AI and cognition points toward a troubling split.

People can perform better with AI while learning less from the experience.

A 2025 meta-analysis covering 228 studies found substantial effects of AI on cognitive and psychological functioning, with generative AI producing particularly large gains in measured cognitive performance. But improvements in metacognition—the ability to understand and monitor one’s own thinking—were smaller and less consistent.

That distinction matters.

A person may produce a better essay, solve a more complicated problem or complete a task more quickly without necessarily becoming better at doing the task independently.

In other words:

AI can improve the result without improving the underlying skill.

Research into human-AI collaboration also produces a nuanced picture. Human–AI teams often outperform people working alone, particularly on creative generation tasks. But the combination does not automatically outperform the stronger individual or the AI system acting alone. In some decision-making situations, collaboration can even reduce performance when people place too much trust in the machine.

The lesson is not that AI collaboration fails.

It is that augmentation is not automatically synergy.

The shortcut trap

Learning has traditionally involved friction.

You struggle with a problem. You make an error. You retrieve information from memory. You try again. You discover why your original approach failed.

That struggle can feel inefficient.

It is also part of how learning happens.

The danger with generative AI is that it can remove precisely the difficult stages that build durable knowledge.

Consider two students facing the same mathematics problem.

The first asks:

“Solve this for me.”

The second asks:

“Give me a hint, let me try it, and then tell me where my reasoning went wrong.”

Both students may eventually reach the correct answer.

But only one has necessarily practiced the reasoning required to reach it independently.

This is the difference between using AI to eliminate effort and using AI to make effort more productive.

When the machine disappears

The most revealing test may be what happens when AI is taken away.

Several studies have found a version of the same pattern: students can perform better on AI-assisted assignments while subsequently performing worse on assessments completed without the technology.

One large study of high-school mathematics in Türkiye reported higher homework performance and less time spent on assignments among students using ChatGPT-style assistance. But when students had to complete later assessments without AI, some groups performed worse than students who had not relied on the technology.

The difference was particularly pronounced among students who substantially reduced the amount of time they spent doing the work themselves.

The implication is significant.

The problem may not be AI assistance itself. The problem may be replacing practice with assistance.

Research has also begun examining what happens to persistence when people receive immediate AI-generated answers. Some experiments suggest that even relatively brief exposure to answer-giving AI can reduce people’s willingness to persist when they subsequently have to solve a problem independently.

That is precisely where the idea of AI as a cognitive crutch becomes important.

What about the brain?

The neuroscience is attracting considerable attention, but it requires caution.

An MIT study using EEG measurements reported differences in brain connectivity among people writing with the assistance of a large language model, along with weaker unaided performance afterward. The findings generated substantial discussion about whether heavy AI dependence could alter how people engage cognitively with writing.

But one study does not establish that AI permanently damages the brain.

The broader scientific picture is still developing, and claims about long-term neurological effects should not be overstated.

What is already easier to observe is behavioral: when people stop generating, retrieving and checking information themselves, they may practice those skills less.

That is a learning problem before it is a neuroscience problem.

The surprising counterexample

The story is not entirely negative.

Some research points in the opposite direction.

In certain writing experiments, people who practiced with AI assistance subsequently produced better unaided work than people who practiced without AI. One possible explanation is that AI can expose learners to high-quality examples and provide a form of feedback that they can later internalize.

This suggests that the critical variable may not be whether AI is present.

It may be what the learner is doing while AI is present.

There is a profound difference between:

“Write it for me.”

and

“Show me three possible approaches, explain their weaknesses, and let me decide which one to develop.”

The technology is almost identical.

The cognitive activity is not.

Coach or crutch?

This may be the most useful way to understand the emerging AI debate.

AI as a coach

You generate an idea.

AI challenges it.

You attempt the problem.

AI gives feedback.

You write the draft.

AI critiques it.

You make the decision.

AI presents alternative perspectives.

Here, the human remains responsible for the core cognitive work.

AI as a crutch

AI generates the idea.

AI solves the problem.

AI writes the draft.

AI makes the plan.

AI provides the conclusion.

The human simply approves the output.

Here, the machine has not merely assisted cognition.

It has increasingly replaced it.

The expert advantage

There is another important distinction: expertise.

An experienced programmer can use AI to work at a higher level of abstraction because they already understand programming fundamentals. They can recognize bad code, test the output and identify when the model is hallucinating.

A novice may not have those safeguards.

The same applies to writing, research, mathematics, medicine, law and many other fields.

Experts can often use AI to accelerate existing knowledge.

Novices risk using AI to substitute for the process of acquiring that knowledge.

That makes the educational question particularly important.

If someone has never learned how to construct an argument, handing them a polished AI-generated essay may produce an impressive document—but it does not necessarily produce an impressive thinker.

Humans have always outsourced cognition

There is nothing completely new about this phenomenon.

Writing allowed humans to externalize memory.

Books multiplied access to knowledge.

Calculators changed arithmetic.

Search engines changed how we retrieve information.

GPS changed how we navigate.

Human intelligence has always been partly distributed across tools, technologies and other people.

Civilization itself can be understood as a vast cognitive system.

But generative AI introduces a new dimension.

Previous tools primarily outsourced memory, calculation or retrieval.

Generative AI can outsource drafting, synthesis, explanation, argument structure and aspects of judgment.

Those functions sit much closer to what we traditionally call thinking.

That is why the stakes may be different.

So, can AI actually make us smarter?

The evidence points toward a qualified answer.

Yes—if we use AI to increase the quality and quantity of our own thinking.

AI can act as:

a personal tutor;
an intellectual sparring partner;
a source of rapid feedback;
a research assistant;
a simulator for testing ideas;
a tool for exploring unfamiliar subjects.

But the same technology can become a shortcut around the cognitive work that creates expertise.

The crucial question is therefore not:

“Did AI help me complete the task?”

It is:

“Would I be better at this task after using AI than I was before?”

That is a much harder test.

The AI TV INFO’s test

Before using AI, ask four questions:

1. What happens when the AI disappears?

If you cannot perform the task without it, the tool may have replaced rather than developed a skill.

2. Am I generating or merely editing?

Producing the initial structure often requires more cognitive work than polishing a finished answer.

3. Can I explain the result?

If you cannot explain why an answer is correct—or where it might be wrong—you may be borrowing intelligence rather than building it.

4. Is AI increasing productive struggle or eliminating it?

The goal should not be maximum convenience.

It should be maximum learning.

The real question is bigger than AI

The debate over artificial intelligence is often framed as a choice between technological optimism and technological fear.

The evidence suggests something more complicated.

AI is neither inherently an intelligence booster nor inherently an intelligence destroyer.

It is a powerful amplifier.

And amplifiers magnify what they are connected to.

Connect AI to curiosity, skepticism, practice and deliberate learning, and it can dramatically expand what a person can explore and accomplish.

Connect it to avoidance, passivity and unquestioned dependence, and it can produce something more deceptive: the appearance of competence without the underlying capability.

That may ultimately be the defining challenge of the AI age.

The question is not whether machines can think.

They increasingly can perform tasks that look remarkably like thinking.

The question is whether we will continue thinking while they do.

Because the smartest use of AI may not be asking it to do the thinking for us.

It may be learning how to make our own thinking better.