In the midst of all debates about how AI affects works, science, the environment and everything else, it is about how large language models affect people who use them directly.
A new study by the MIT media laboratory implies that the use of AI tools reduces brain activity in some way, which is understandably alarming. But I think that is just a part of the story. The way we use AI, like any other piece of technology, is what really matters.
This is what the researchers did to test the effect of AI on the brain: they asked 54 students to write tests using one of the three methods: their own brains, a search engine or an AI assistant, specifically chatgpt.
In three sessions, the students stayed with their assigned tools. Then they exchanged, with the users of AI without tools, and the users who are not of Tool use.
The EEG headphones measured their brain activity at all times, and a group of humans, in addition to an especially trained AY, obtained the resulting trials. The researchers also interviewed each student about their experience.
As expected, the group that depends on their brains showed the greatest commitment, the best memory and the greatest sense of property about their work, as evidenced by how much they could cite them.
Those who used the at first had a less impressive retirement and brain connectivity, they often could not even quote their own trials after a few minutes. When writing manually in the final test, they still have a lower performance.
The authors are careful to point out that the study has not yet been reviewed in pairs. It had a limited range, focused on the drafting of trials, not on any other cognitive activity. And the EEG, although fascinating, is better to measure general trends than identifying exact brain functions. In spite of all these warnings, the message that most people would take away is that using AI could make you fool.
But I would rethink that to consider if maybe AI did not go as much as we can choose to think. Perhaps the problem is not the tool, but how we are using it.
AI brains
If you use AI, think about how you used it. Did you get it to write a letter, or maybe make a rain of ideas about some ideas? Did you replace your thinking or support it? There is a big difference between the subcontracting of an essay and using an AI to help organize a messy idea.
Part of the problem is that “ai”, as we refer to him, is not literally intelligent, just a very sophisticated parrot with a huge library in his memory. But this study did not ask the participants to reflect on that distinction.
The LLM use group was encouraged to use AI as they considered convenient, which probably did not mean reflective and judicious use, just copy without reading, and that is why it is the context.
Because the “cognitive cost” of AI can be less linked to its presence and more with its purpose. If I use AI to rewrite a boiler email, I am not decreasing my intelligence. On the other hand, I am releasing bandwidth for things that really require my thinking and creativity, such as devising this idea for an article or planning my weekend.
Of course, if I use AI to generate ideas, I never bother to understand or interact, then my brain probably takes a nap, but I use it to rationalize tedious tasks, I have more intellectual capacity for when matters.
Think like that. When I grew up, I had dozens of telephone numbers, addresses, birthdays and other details of my friends and memorized family. Most had written somewhere, but I rarely needed to consult it for those who were closer. But I have not memorized a number in almost a decade.
I don’t even know my own fixed number of memory. Is it a sign that I am getting more silly, or just evidence that I have had a cell phone for a long time and stopped needing to remember them?
We have downloaded certain types of retirement to our devices, which allows us to focus on different types of thought. The ability is not to memorize, it is to know how to find, filter and apply information when we need it. Sometimes it is known as “extrayligence”, but in reality it is only applying the brain power where it is needed.
That does not mean that memory no matter. But the emphasis has changed. As we do not make students practice the long division by hand once they understand the concept, one day we will decide that it is more important to know how a good letter is seen and how to ask an AI to write one that writes it by line from scratch.
Humans are always redefining intelligence. There are many ways to be intelligent, and knowing how to use tools and technology is an important intelligence measure. In a moment, being intelligent meant knowing how to attack Flint, making Latin decleions or working in a sliding rule.
Today, it could mean being able to collaborate with machines without letting them think about you. Different tools prioritize different cognitive skills. And every time a new tool appears, some people get scared that it will ruin us or replace us.
The printing press The calculator. The Internet. All were accused of making people had lazy thinkers. Everything turned out to be a great blessing for civilization (well, the jury is still on the Internet).
With the mixture, we are probably tilting more in synthesis, discernment and emotional intelligence, the human parts of being human. We do not need the type of scribes that are only good to write what people say; We need people who know how to ask better questions.
Know when to trust a model and when to verify twice. It means converting a tool that is capable of doing the work into an asset that helps to do it better.
But none of that works if you treat AI as an vending machine for intelligence. Hit on a warning, wait for brilliance to fall? No, this is not how it works. And if that is all you do with that, you are not becoming more silly, you never learned to keep in touch with your own thoughts.
In the study, the ownership of lower rehearsals in the LLM group was not just about memory. It was commitment. They did not feel connected to what they wrote because they were not the ones who wrote. That is not about AI. It is about using a tool to omit the difficult part, which means omitting learning.
However, the study is important. It reminds us that the tools shape thought. It pushes us if we are using AI tools to expand our brains or avoid using them. But to affirm that the use of AI makes people less intelligent is like saying that calculators made us bad in mathematics. If we want to maintain our acute brain, perhaps the answer is not to avoid AI, but be reflective about using it.
The future is not a human brain versus ia. These are humans who know how to think with AI and any other tool, and avoid becoming someone who does not bother to think at all. And that is a test that I would still like to happen.
