Retour Ă  la liste
Technologie et réseaux sociaux Fact-check publié le 21/09/2025

🟧 ChatGPT transforme nos cerveaux et nous rend bêtes

Le post original

ChatGPT is silently changing our brains. MIT reveals how to save them: We've all felt it: You get used to doing something with AI. Then you try doing it on your own. And it's hard. Really hard. MIT's MediaLab just did one of the first studies into how using AI affects us. Over 4 months, they used brain scans to study 3 groups of people doing a writing task. The different groups wrote: On their own. With Google Search. With ChatGPT. The study showed using AI can lead to: 1/ Memory loss 83% of ChatGPT users could not recall a single sentence they had written just minutes earlier. People writing without AI remembered fine. 2/ Less brain power 47% decrease in brain connectivity in ChatGPT users. This continued even after they stopped using AI. 3/ Less learning 33% decrease in the mental effort that leads to learning. The price you pay for finishing faster. The solution? It's not to throw AI out the window. The research showed you can have: Stronger memory Better brain activity Higher quality work All while still using AI. The key: Don't use AI at the start. Start with your own thinking. Bring in AI later to improve or expand. Use AI as an editor or assistant, not a substitute. Our brains have been here before. We remember less from: Typing vs. writing by hand Taking photos vs. just looking Using sat nav vs. navigating ourselves The thing that makes the difference: Being intentional. Use your brain to think, learn, and create. Then let AI help with the rest. Remember: When you hand over your thinking too soon, you aren't just finishing faster. You're rewiring your brain. What’s one task you won't let AI take over? I'd love to hear below. ♻️ Repost to help others protect their brains. 🔔 Follow me (James Ware) for more like this.
Original post image

Le fact-check

🟧 True, but the author either inflates the claims or distorts the conclusions of the study. The MIT Lab did publish such a study http://bit.ly/4ml1jXD as a preprint (not yet peer-reviewed). The study is not longitudinal and cannot be used to conclude to a temporal effect, let alone a persistent one. 1) Memory loss: no such vocabulary exists in the study. It tackles “recall problems” for that specific task of writing short essays. Subjects in the Search-engine and Brain-only groups also experienced recall problems, but less so. The percentage mentioned is from Session 1 only. It drops to 33 % at session 3. 2) Less brain power: no such vocabulary exists in the study. It mentions “cognitive engagement” and “cognitive load”, both lower in the LLM group, having opposite effects on the quality of learning. 3) Less learning: the study did not involve learning. This is a leap the author makes. For less black and white comments, we suggest https://bit.ly/4pncyRU and reading the article. However, don’t ask an LLM because it will only read page 3 https://bit.ly/4mo5SQG. It is strange to us that the author does not mention that fun fact. Disclaimer: no AI was used nor harmed in writing this comment.

Historique

2 Ă©tapes
18 sept. 2025 • 06:53

Post Reçu

21 sept. 2025 • 07:15

Publication

🟧 True, but the author either inflates the claims or distorts the conclusions of the study. The MIT Lab did publish such a study http://bit.ly/4ml1jXD as a preprint (not yet peer-reviewed). The study is not longitudinal and cannot be used to conclude to a temporal effect, let alone a persistent one. 1) Memory loss: no such vocabulary exists in the study. It tackles “recall problems” for that specific task of writing short essays. Subjects in the Search-engine and Brain-only groups also experienced recall problems, but less so. The percentage mentioned is from Session 1 only. It drops to 33 % at session 3. 2) Less brain power: no such vocabulary exists in the study. It mentions “cognitive engagement” and “cognitive load”, both lower in the LLM group, having opposite effects on the quality of learning. 3) Less learning: the study did not involve learning. This is a leap the author makes. For less black and white comments, we suggest https://bit.ly/4pncyRU and reading the article. However, don’t ask an LLM because it will only read page 3 https://bit.ly/4mo5SQG. It is strange to us that the author does not mention that fun fact. Disclaimer: no AI was used nor harmed in writing this comment.

Vous avez une question, une remarque ou une suggestion ? Contactez-nous, nous vous répondrons au plus vite !

Nous contacter
Posez votre question Ă  VeraVera