Thread 508023316

In first of its kind, a preliminary MIT study of brain scans shows that using LLMs results in 47% reduction in brain connectivity.

>Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels.
When researchers forced ChatGPT users to write without AI, they performed worse than people who never used AI at all.

>LLM users also struggled to accurately quote their own work.
83.3% of ChatGPT users couldn't quote from essays they wrote minutes earlier.

Sample size: 54
Link to the paper: https://arxiv.org/abs/2506.08872

>>508023316 (OP)
47%
holy fuck

>>508023316 (OP)
The only people surprised by this are ones who rely on AI to think for them.

>>508023316 (OP)
>people outsource their brain to new tools
okay and so what?
we've been doing this since the invention of the written word

>>508023316 (OP)

Absolute bullshit tier "research"

>>508023316 (OP)
In general, relying too much on any technology makes one retarded.

>>508023316 (OP)
Working as intended.

>>508023936
when you outsource your thinking, what did you think would happen? you'd somehow magically be good at thinking? nothing in life works that way.

>wax tablet users consistently underperformed at remembering
>printing press users consistently underperformed at calligraphy
>abacus users consistently underperformed at mental arithmetic

>>508023316 (OP)
brains don't exist
brain activity doesn't exist
scientists are liars

Yeah no shit
You aren't using your brain you're using software that thinks for you
The more you rely on technology the weaker you become which is why we need irl pol where we can get together and argue in a real physical space

>>508023316 (OP)
Chat GPT is helping me start a business from scratch. Been really helpful.

The emperor told you to not use AI, you did listen right anon?

>>508024132
>we've been doing this since the invention of the written word
Bad analogy, write and read expand the brain, contrary to nawvpAI

>>508023316 (OP)
>Sample size 54
Were you included in the study?

>>508023316 (OP)
Correlation is not causation. Perhaps retarded people like AI

>>508023316 (OP)
>hmm could it be that retarded people rely on AI to think for them?

>no, AI is making them retarded

>>508023316 (OP)
I've been testing chatgpt for years, the only difference is does is for the lazy to be more lazier.

>>508023316 (OP)
the simple act of often glancing at a gps for directions kills the entire mental process and prevents the driver from ever entering a phase of 'deep practice'

people who drive this way have near zero driving ability. whereas somebody using an old fashioned map or even a garmin that they use for occasional routing like a cab driver develops a much larger behavior cycle and decision process that is really complex. They're optimizing possible routes between two points with a lifetime of knowledge about street level texture.

following instructions step by step is for retards and makes retards and is retarded.

>>508023316 (OP)
llms are a tool. if youre a retard, youll stay retarded. if youre smart, its a great research assistant
if you dont know how to use tools effectively, youre retarded

technology makes you dumber, eventually causing everything to collapse because you need intelligent beings to maintain a modern society. funny how its built to destroy itself in the long term.

>>508023316 (OP)
BREAKINGGGGG!!!!!111

>>508023316 (OP)
Writing essays is a ridiculous waste of time for almost everybody
I've never written shit as an adult
What use is an essay for an IT person? Or a programmer? Or a doctor? Or even a lawyer?
It's good for history and english majors and fuckers who wanna go into academia

>>508023316 (OP)
Sounds like made up bullshit to me. The existence of this papers isn't the psyop, but the inorganic pushing of this narritive is.

>>508027389
"Science" is a joke and this trash was likely written using jeetPT itself to add to the irony.

Ai is fine it will help the peasants peasant harder

>>508023316 (OP)

it's not just AI doing it, either

>>508025549
>Don't use your eye glasses it makes you retarded

>>508027614
>>508027614
>What use is an essay for an IT person? Or a programmer?
You must belong to the my code is self documenting camp

>>508023316 (OP)
Since I'm not reading this right now, may I ask if they controlled for the proclivity of already dumb people relaying on LLMs more than the mean
I could readily believe it makes people retarded without any studies at all but if you polled who uses swimming aids you would find mostly people who can't swim rather than those who could and their skills deteriorated as a consequence
unless I'm misunderstanding the title, I'm really tired

>>508027810
>documenting code
Yeah writing what a line does either in the code itself or a line or two in a text document surely requires 2000 words of researching the dumbest shit imaginable

>>508027826
>asking questions

some topics are better off left un-explored, anon

>>508023316 (OP)
If I was going to write a paper I would need to do research in a library or later on the internet but now ChatGPT can be used to quickly collect relevant information into useful lists with context. I've personally learned a lot by doing research through this tool. I retain the information that I learn and modify my understanding of topics based on new information I gather through this tool.

I believe what this "study" with it's incredibly tiny sample size of 54 is actually showing is that a lot of people are being lazy cheaters. IF you are tasked with a homework assignment on some gay boring subject and you just have chatGPT crap out the whole report and you barely even read the output because it's so fucking lame and gay, then of course you won't retain the information because why would you? Who gives a shit? Homework is busywork about stupid shit nobody gives a fuck about. In fact we generally hate it. Do you know how hard it is to retain information that you hate learning? It goes in one ear and out the other and doesn't leave much of an impression.

I retain all the information from all of my research because I did it out of curiosity. I wanted to know and now I know. I wasn't forced to do research to get a grade. I never hated doing the research. It was a joy. I enjoyed it and I remember it well. I'll retain that info and use it in my life and in future discussions. I became even smarter and more educated.

>>508023316 (OP)
Patch: critically read, analyze, edit, and revise the output. Do not just 'direct' the system. You too must do heavy lifting!

>>508023316 (OP)
Did they account for lower IQ users being more likely to use LLMs to perform above their intellectual limits using tech, compared to high IQ users not needing it?
>muh fMRIs
Yeah nah garbage. Fucking reading tea leaves mate.

>>508023316 (OP)
How is it any different from googling something you want to know?

>>508023316 (OP)
I just make chat gpt tell me bad fan fiction stories.

Study is flawed for a few massive reasons. It's a seriously bullshit study. all these flaws significantly undermine the study's bold claims about AI causing cognitive debt.

>there were 54 people in the study in 3 different sessions, but only 18 participants among them completed session 4 where they switch groups.
>while EEG can measure brain activity, it has poor spatial resolution and cannot definitively establish the cognitive processes the researchers claim to measure. the interpretations of brain connectivity patterns are largely speculative.
>human teachers scored essays without knowing group assignments, but the AI judge disagreed significantly with human teachers. This is suggesting fundamental issues with measurement validity. The study notes teachers could often identify AI generated essays, indicating the blinding wasn't effective.
>the study analyzes numerous brain regions and frequency bands without adequate correction for multiple comparisons, increasing the risk of false positive findings.
>participants had varying levels of prior ChatGPT experience , but this wasn't adequately controlled for in the analysis.
>different essay topics across sessions could influence results independently of the writing tools used.
>participants knew they were being studied and that their brain activity was being monitored, which could significantly alter their natural writing behavior.


(1/2)

>>508026824
>thinking machine users consistently underperformed at cognitive tasks
Brainchads, report in!

>>508028299
>study interprets reduced brain connectivity as "cognitive debt" without establishing that less connectivity necessarily indicates impaired cognition. Reduced activation could also indicate efficiency or automaticity.
>study assumes that observed neural differences directly cause the behavioral differences, but correlation doesn't establish causation.
>extensive data collection across multiple brain regions, sessions, and measures, there's risk of selectively reporting significant findings while downplaying null results.
>essay writing was tested. the cognitive effects might be completely different for other writing tasks like creative writing, technical documentation, or professional communication.
>study lacks a proper control group that writes essays under normal conditions without time pressure or brain monitoring.
>study doesn't establish adequate baseline measures of participants' writing ability, cognitive function, or brain activity patterns before the interventions.
>despite claiming to study "cognitive debt," the study doesn't follow participants long term to see if any effects persist or recover.


(2/2)

>>508028131
These days a ChatGPT query is infinitely more useful than a google search result since the google search results only give you advertisements now. You are lucky if there are 3 actual websites at the bottom of 4 pages of advertisements.

>>508028131
Google is straight up garbage nowadays

>>508028299
>>508028347
out of all of the issues outline here, I think this one is the most damning.
>there were 54 people in the study in 3 different sessions, but only 18 participants among them completed session 4 where they switch groups.

>>508028005
True, but at the same time not true. Sometimes depending on your job you must learn things you don’t like simply because they’re required. An engineer may want to focus on building but safety procedures and international health standards are part of the job.
I don’t mean stupid things like STEM being forced to take intersectional ethics classes taught by some Karen, I mean important aspects of your job that are simply boring and impossible to enjoy

>>508028347
>Reduced activation could also indicate efficiency or automaticity.
mum says my brain is efficient

>>508023316 (OP)
I definitely feel dumber after getting a smart phone, even with a calculator in my pocket i can barely do basic math in my head anymore

>>508028481
Maybe I'm a weirdo but I actually get angry when people try to teach me something I don't think is useful. And I've actually regretted this mindset a time or two. In high school I did terrible at my Algebra class because my teacher was a 75 year old boring woman with a weak voice who did nothing but give out 10 pages of homework every single day. I didn't see a use for algebra in my life so I felt like this was a deeply offensive waste of my time and mental resources so I refused to do any of my Algebra homework and in class I would defiantly tell the old bitch that I didn't do my homework again and that I never would. Later in life when I wanted to get into programming, I actually started to need some Algebra, but the problem was clearly that my teacher sucked 1000 antique cocks from 1929 and she never made any effort to explain what her gay algebra was ever going to be useful for.

>>508027319
I bet since we gave up on memorising poetry our memory's got much worse

>>508028358
Bro for your own sake install ublock origin. Just do it. Holy shit I cannot imagine the hellscape you live in.

>>508023316 (OP)
Chat GPT is retarded anyway.

>>508027541
This. Every day there are reports of critical Japanese infrastructure failing, ostensibly for lack of competence. I think the Amish may have achieved the peak level of sustainable human development.

>>508028913
same here, nearly same story

>>508023316 (OP)
>people who write essays better than people who haven't been writing them that way for months.
whoa.

>>508023316 (OP)
Yeah, and people who use calculators also have a reduction in arithmetic ability.

>>508027614
if you can't articulate your ideas it is generally because you don't have any

Im in my last semester of undergrad and the faculty and institution as a whole have no idea how to counter or integrate it appropriately. Fucking whatever. These faggots make way too much anyway.

>>508023316 (OP)
I played with ChatGPT for a couple of days, quickly got bored of it and haven’t used it or any other AI since. You ask it a question and it always responds with the same soulless:
>there are many diverse opinions on [topic], including
>bullet pointed list
It’s nothing like having a debate or even a conversation with an actual human. I don’t know what people see in it.

I finally turned off autocorrect and it feels so good nig nog digger dog hitler

>>508027319
@grok explain

>>508023316 (OP)
Only for npc
I'm literally becoming smarter
And I also don't use it

>>508029553
I use the Brave browser. But I can still see that google is giving me 5 results instead of 5 million.

>>508023316 (OP)
Computer, generate me a 49 foot tall futsnari anthro pony. Huge breasts, micro bikini, beach background, sand, ocean, sky, clouds, phototralistic style

>>508023316 (OP)
In my experience obsession with / looking highly upon "AI" is a terrifically accurate predictor of lower intelligence. Not in every case of course, but enough to make it significant.

>>508031265
>I use the Brave browser
It's very cool, anon but that's nothing
LibreWolf, integrated uBlock and Scriptmanager