"While LLMs offer immediate convenience, our findings highlight potential cognitive costs. Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI's role in learning."
https://www.media.mit.edu/publications/your-brain-on-chatgpt/
@strypey A bit of a pet peeve, but that study rubs me the wrong way. Not the results, but that we had to have a study to confirm the results.
It'd be like having a study where one group fixes their car for three months, and a control group pays a mechanic to fix their car instead. Of course the people fixing their car are going to be more mechanically proficient in the end, and if you measure brain activity, fixing a car will show more brain activity than sitting in the waiting room of a garage.
/end rant
@nate
> if you measure brain activity, fixing a car will show more brain activity than sitting in the waiting room of a garage
Sure. But if you swap the groups over, you'd expect that the group now fixing their own car would show the same level of brain activity as the previous group, who are now waiting for the mechanic. What the study shows is that the erosion of brain activity during LLM use persists, even once they're asked to write their own essay unaided.
@nate I still seems to me like you're missing the point, but I'm not sure how to explain why I think so more clearly. Maybe read the full paper (again)?
Maybe the problem is not your logic, but your premises. Particularly the idea that having a mechanic fix your car would lead to a persistent, reduced level of cognitive activity, in the way that using LLMs to write essays seems to. I don't see why it would.
Are you aware of any evidence that supports this proposition?
