Skip to main content

Study: Using AI for 10 Minutes Could Make You Lazy and Dumb

Using AI for just 10 minutes can make you a little lazier and dumber. Researchers at Carnegie Mellon, MIT, Oxford, and UCLA found that using AI for even 10 minutes may have a shockingly negative impact on your ability to think and problem-solve.

Released on 05/07/2026

Transcript

Using AI for just 10 minutes

can make you a little bit lazier

and a little dumber it seems.

Researchers at Carnegie Mellon University,

MIT, Oxford University and UCLA

found that using AI for just 10 minutes

can have a shockingly negative impact

on your ability to think and problem solve.

The researchers tasked people with solving problems

like simple fractions and reading comprehension

through an online platform that paid them for the work.

Some were given access to an AI assistant

capable of solving the problem autonomously too.

Their study found that when the AI helper was taken away,

people were significantly more likely

to give up on the problem or get their answers wrong.

This suggests that although AI might boost productivity,

it could be at the expense

of developing foundational problem solving skills.

And MIT professor involved with the work told me that

the aim was to take broader concerns

about long-term human to AI interaction

and study them in a controlled experiment.

He says the takeaway is not that we should ban AI

in educational workplaces, but that we should think about

how AI can be designed to help humans best.

For example, it may be necessary to design AI tools

so that they sometimes prioritize a person's learning

over just solving problems for them.

Putting too much faith in AI is especially problematic

when the tools don't behave as you expect.

Agentic AI systems can be particularly unpredictable

because they do complex things in ways

that can reduce surprising errors.

AI companies are already thinking about

the more subtle effects that models can have on users.

OpenAI, for example, previously toned down

the so-called sicker fancy in some of its models.

I recently learned the danger

of offloading critical thinking to AI myself

after letting OpenClaw troubleshoot my Linux machine

in a way that left it unable to boot.

Perhaps instead of trying to solve the problem for me,

that AI system could have pause

and taught me how to fix the issue for myself.

For more like this, sign up to my AI lab newsletter

at Wire.com/newsletters.