Paul Cantrell
❝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.❞
Hell of a research abstract there, via @gwagner: https://fediscience.org/@gwagner/114690366530883451
Paul Cantrell
“Using LLMs makes you stupid” is such an emotionally appealing conclusion that I’m going to consciously work — this post is my public commitment! — not to read •too• much into this one study. As the OP says: small N, not peer reviewed, etc.
But I •will• immediately heed the authors’ “concerns about the long-term educational implications of LLM reliance.” However the research shakes out, it seems like good practice to double down on helping students keep their own brains awake and active regardless of the technology available to them.
Paul Cantrell
Whatever an EEG means to functional intelligence, that there was any measurable difference at all is certainly eyebrow-raising. Taking the conclusions as established fact is obviously unwise, but ignoring the possibility doesn’t seem wise either.
Das
@inthehands @gwagner indeed! Small n and I have no idea how (well) the EEG reports count - hopefully the appropriate peers will address.
But I suppose this buildup of 'cognitive debt' would lead to cognitive delinquency...
Paul Cantrell
@venya
Exactly.
Venya (he/him/dude) 🇺🇦
I mistrust something which so perfectly reflects my own gut feelings.
Paul Cantrell
More from @grimalkina to the effect of “maybe don’t read too much into this paper’s conclusions:”
Paul Cantrell
I see @datarama has a similar reaction about a study that says something we want to hear.
Re that second remark, “The *point* of AI is to make cognitive abilities irrelevant,” well, that’s the thing that’s up for debate right now. Making cognitive abilities (and thus labor, and people) irrelevant is very much the marketing pitch of the hype bubble. That pitch is about cost cutting, and fear and intimidation, an intoxicating mixture of fantasy and terror.
It’s not the only vision of AI, however.
Paul Cantrell
There has long been a quiet debate, currently drowned out but still very much happening, about human replacement vs human augmentation. Think of Gary Kasparov remarking years ago that he thought chess played by humans with computer assistance could be a far more interesting game than either human-only or computer-only chess.
Here’s an argument for augmentation, and not how far it diverges from the current hype despite it being written from a very AI-friendly point of view:
Paul Cantrell
When I say that I wish the bubble would burst so we could start having actual conversation, this is the kind of thing I’m talking about.
In the meantime, not being a researcher myself in any kind of position to critique the paper, I’m happy to just look at it and cherry-pick the conclusion that I think is unequivocally helpful regardless: we all (students in school, students of life) should cultivate active, engaged minds that work with ideas by •doing• and •creating• instead of simply passively receiving.
Paul Cantrell
@nataliyakosmyna Oh, hi! Yes, I did not mean to imply “LLMs make you stupid” was •your• language at all; I’ll edit to clarify. That’s just the confirmation bias bell it rings in my own brain, and thus the reason I’m treating the results with caution.
Results aside — I truly am not equipped to judge them! — your research question seems a worthy and important one, and I’m glad you’re asking it.
nataliyakosmyna
@inthehands we do call this on-going phenomena “cognitive debt”.
nataliyakosmyna
@inthehands just as a note, we did not use the language “LLM makes you stupid” in the paper. It is written in an academic manner, though we did add TL;DR section as well as a summary table for an LLM, as the final document is 206 pages (with appendix). But we did show the reduced neural connectivity and a lot of other issues LLM group faced. If you are interested, our summary of the paper is here (authors of this paper): https://www.linkedin.com/posts/nataliekosmina_mit-ai-brain-activity-7340386826504876033-X45W?utm_source=share&utm_medium=member_ios&rcm=ACoAAANkfbABvU568kU63aYOiOdVABVfyyA2Trs
Paul Cantrell
Hey, the paper’s first author, @nataliyakosmyna, in on Fedi! Let’s welcome her, and take a moment to note what she says about the study:
Paul Cantrell
@nataliyakosmyna
Oh, FYI: I’m not getting any audio from that video. Maybe a LinkedIn problem and not on your end, but worth a check.
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.