>da Chinese have fallen prey to techno-supremacy
>da west has fallen :(

>>25612920
The last literate generation was 2 generations ago, about 10% of adults are capable of a collegiate reading.

Welcome to the aristocracy, may your patron be wealthy and shower you with gifts and praise.

>>25612920
nig wht u mean litreate?
lol chinks

>>25612924
>The last literate generation was 2 generations ago, about 10% of adults are capable of a collegiate reading.
fun fact: a phd colleague of mine is a proud non-reader of books

>>25612930
can he read his diploma?

Maybe I've used too much AI, but I'm having a hard time making sense of the chart. I assumed at first glance the x-axis was supposed to be year, but that doesn't appear to be the case on closer inspection? Is "Homework score..." a label for the x-axis, or is it a caption for the chart? If the x-axis is the average homework score, what does the y-axis represent? Is "homework score" different from "exam score?" Does anyone have a link to the article? Or maybe I should just ask ChatGPT.

>please make it count
For myself I absolutely will. For the third worlders and shitskins who’ve flooded my society? Lol no thanks. This little jewish experiment of a civilization needs to fail as soon as possible so they can all starve to death.

>>25612920
interesting how the AI user score only significantly diverges in the high homework score range

>>25612958
>Is "homework score" different from "exam score?
Of course.
Meaning, the average homework grades are increasing, because they are using AI to do it.
However, when they sit for exams, they do significantly worse.
People who don't use AI have about the same grades in homework and exams. But people who do, do great in homewok but bad in classroom exams. Worth noting, people who get the highest scores in homework, do worse on tests, because they let the AI do everything.
(I thought the x axis was years and didn't read it until you point it out...)

>>25612920
>make it count
How? By reading books? I’m already doing that, desu.

>>25612920
What's the Y axis?

>>25612924
Only 2% have Level 5 literacy, defined as:

>Adults can search for, and integrate, information across multiple, dense texts; construct syntheses of similar and contrasting ideas or points of view; or evaluate evidence based arguments. Adults understand subtle, rhetorical cues and can make high-level inferences or use specialised background knowledge.

This is a basic requirement for substantial free will, meaning ability to research, interrogate and change your beliefs and understanding of the world.

Your 10% are the level 4 literates:

>Adults can integrate, interpret or synthesise information from complex or lengthy texts. Adults can identify and understand one or more specific, non-central idea(s) in the text in order to interpret or evaluate subtle evidence-claim or persuasive discourse relationships.

This means that they can understand and regurgitate a single complex text. Basic "evaluation of evidence-based arguments?" That belongs to the 2%.

>>25612920
This chart makes zero sense.

>>25613481
Filtered

>>25613218
>What's the Y axis?
Exam score. This anon explains it: >>25612998. The AI-using students who did 30% better than average on the homework did 35% worse on the test. When using AI, doing well on homework has an inverse correlation with test performance (vs non-AI students, where homework score tracks with test score).

The point of homework is to develop skills. Tests measure those skills. Using AI to do your homework is cheating yourself out of developing the skills you need for tests.

>>25613679
>Filtered
Yeah, the math teacher used to filter me if i didn't label the X and Y axis. Dumbass.

btw the left-hand side is less clear, it seems likely that those are students who are only using AI to enhance a normal homework practice. So they are still developing skills from the homework scores are boosted slightly due to AI's help.

>>25613686
The label is there, dumbshit, it's just at the top. You are filtered because you're sub literate.

what about for the people who just use ai as a tutor and not as the givemeanswers machine? seems reductive to only have the cheaters versus the noncheaters

>>25613690
Go do that in a class and call the professor a dumbshit illiterate next, then post proof, once you grow a pair.

what are the vertical axis numbers

found the preproont
https://cepr.org/publications/dp21577
Using 30 months of panel data on 26,811 Chinese students in grades 7--12, we study how generative AI affects homework productivity and learning. The data combine monthly closed-book exams, high-school and college entrance exams, and homework scores and completion time across nine subjects. We exploit staggered AI adoption in a difference-in-differences design. AI adoption raises homework scores by 18% and reduces completion time by 30%, but lowers monthly exam scores by 20% within six months. High-stakes entrance-exam scores fall by 18 and 24%, with the full penalty emerging only after about two years. The losses are largest in social science subjects, followed by STEM and languages, and are especially large for junior students, high-achieving students, and boys. The learning losses are concentrated among roughly 80% of AI users whose behavior is consistent with homework outsourcing, as indicated by exceptionally short homework completion time coupled with high homework scores. AI users who maintain similar homework completion time as non-AI users experience small learning losses.

reading thru though there isn't any effort to differentiate the self-selection that goes into using the AI

also somehow the paper analyzes a fixed cohort reporting AI adoption in October 2022 when ChatGPT was launched that november. idk what AI they're even referring to that came out before then, much less whether all the models of AI should be considered the same for this

>>25613694
Still seething about this? Holy shit
You had terrible teachers who completely failed to prepare you for life. Be less retarded next time. You can do it, I believe in you.

>>25613691
It's funny how the people who doom about AI making everyone stupid are themselves morons who only use AI as the givemeanswers machine and thus are incapable of even considering the usage of AI as a tutor. The fact of the matter is that lazy morons will continue to be lazy morons with or without AI, while people who actually want to learn will be able to use AI to learn slightly faster.

>>25613679
>Exam score
Sorry I saw it after I posted. To me it looks like the students who figured out how to cheat the best on homework tried to rely on that to pass knowing they would scrape by with flunked exams. I'm not sure it's just laziness, there has to be some calculated neglect involved.

>>25613713
Doesnt china have a cheating culture anyway?

>>25614554
like America doesn't

>>25612920
> never used AI x using AI scores go in parallel up to ~ 103 pts
> Using/not using difference 1-2 pts for non-users
> "GLM, do my homework no mistakes" laz boyz tank exams
> Working together with AI actually makes students learn
News at five

>>25614640
chinese university students literally attack their own campuses when the teachers there implement ANY anti-cheating measures: cheating is accepted as the norm to the point where trying to restrict it at all results in widespread violence. kind of like when a cop shoots a nigger in the US. chinks worship cheating.

>>25614666
a much healthier perspective. Cheating wins outside the artificial membrane of academia as well. Best not to lie to oneself, and be prepared for real world conditions.

>>25612920
Shit chart without seeing how many pupils are at each score level.

>>25614643
>Working together with AI actually makes students learn
Patently false. There are smaller but still noticeable losses associated with atypical users too. There is no circumstance in which AI improved the learning outcomes.
>>25614720
N=26,811 if you're concerned about low counts.

>>25614905
>Patently false. There are smaller but still noticeable losses associated with atypical users too. There is no circumstance in which AI improved the learning outcomes.
You didn't even read the paper, retard
>Our results may appear to contrast with Kestin et al. (2025), who find that Harvard first-year physics students learn more in less time with a specialized AI tutor than with in-class active learning. However, their result is not inconsistent with ours. They demonstrate that, with a well-designed tool and highly motivated students, AI can enhance learning efficiency. We show that, under realistic conditions in which teenagers choose their own tools, the dominant pattern is not tutoring but homework outsourcing and a large AI learning penalty.

>>25615129
it's fucked

>>25614681
I once had a Chinese student ask me why Americans finishing college were smart when they did so badly in highschool. It's because college in China is just cheating and bribes and slacking off, apart from their two best schools, where some students achieve but it's still 90% cheating and bribes and slacking off. That's not what I said though. What I said was oh we specialize sooner which is very appealing to them because they have legitimately too much homework and anything that sounds like it would reduce that is a good idea to a Chinese high school student.

>>25615379
A very diplomatic answer. You prevented a minor international conflict.

Ever read Bo Yang's dirty chineseman?

>>25615129
access to good tools and a structured environment implies a very two tiered future. Somewhat like knowing how to read was a major advantage.

>>25615381
No, but I just ordered a used copy off Amazon because you mentioned it. Sounds like a trip down memory lane is coming my way soon.

China is weird because you meet a lot of good people there and feel bad that they have to live in China, but also you think the other Chinese you haven't met are reaping what they sow and fuck em anyway. The cheating, conniving, needless jingoism, spitting, pollution, disorder, huge smiles invites into people's homes, noodle eating aggression, consumerism, generosity, curiosity, that it's considered bad manners to hit your opponent when they are stunned in Street Fighter in an arcade where every machine is an emulator, pigs being killed on a sheet of untreated plywood in a wet market, beggers who give happily give you change of you ask.

>>25615396
>noodle eating aggression
I'm sorry what?

>>25612920
RACE WAR

>>25612920
You luddites are so annoying.
Not having to think is great. You can outsource all the annoying shit to AI and then play games as much as you want but somehow this is bad.

Beyond a certain level education is dysgenic in women. We should look at the entire system itself as antifeminine, some forms of it have positives for men but that's disputable.
the main beneficieries appear to be hamosexually inclined men.

However did we end up like this?
It steals away the best years of your life and rewards you with useless skills and debt.
It's so severe in women because by the time the system finally spits them out they've been robbed of the children they rightfully deserve to have. and the smarter she is the later it's spits her out, then the more children will have been stolen from her.

This is demonstrably evil.

>>25615471
*homosexually inclined men.

>>25613691
I find that it still gets shit wrong too often and draws from cliche sources.

Any fucking question about the medieval era has 1/3rd of the answer dedicated to the black death, and so many questions use the line about city air making you free. It's frankly annoying.

>>25615471
Sorry, meant this for the write your thoughts thread.

>>25615129
>Our results may appear to contrast
>contrast
I didn't expect /lit/ to be this illiterate. Also the Harvard study talks about "specialized AI tutors" and not generic LLM use for homework.

>>25615515
holy shit, learn to read

>>25613679
>>25612998
It still makes no sense. Why are there students using AI who do badly on their homework? Are they using the worse models? And then, of the students using AI, those who perform worse on homework perform better on the exam.

But you would expect the opposite: that those who score well on homework with AI would be smarter than those who score badly on homework with AI, since they know how to choose the better technology, how to write superior prompts, how to validate the output better, and so on...

The alternative is that "Using AI" means at least some portion of their homework was completed by AI, and the homework score within the AI cohort is a proxy for the proportion of their homework that was completed by AI, so that those who rely on it more score worse on exams.

If that is the case, though, would it not be better to have the X-axis as "percent of homework generated by AI" and the y-axis be the exam score? Since then, rather than plotting a proxy, we would be looking at what we are interested in: namely, the effect of AI usage dependence upon exam scores.

>>25612920
YOU!

...are the last literate generation. The Chinese have had script for more than 6000 years. They'll be fine.

YOU!

...just have to worry about your Capitalization at the beginning of sentences, punctuation, and a neater way to convey the meaning of 'fucked' without typing 'fucked', because fuck you me, I do nót want to be involved in fuckery, and neither doeth thee.

>>25614905
>N=26,811 if you're concerned about low counts.
I'm not concerned about low counts. I'm concerned about distribution. If most of the AI users are in the middle, but most of the non-users are on the left, then most AI users could still be doing better than most non-users.

Arithmetic means are not useful for frequency analysis like this.

>>25616446
>I'm concerned about distribution
The paper's abstract describes that at least "80% of AI users" fell into the category of high scores for homework and low scores in exams. So I'd expect the distribution to skew heavily to the right.
>If most of the AI users are in the middle, but most of the non-users are on the left
Even if that was the case, that would be because of sampling bias and the graph would still highlight a gap between in-class and out of class performance. An AI user is always expected to do worse than a non-AI user if they produced the same quality of homework, which seems to be the overall message of this paper. Reduction of time spent on homework appears to have the same utility as reducing the time spent in the gym during a workout.

>>25616490
>The paper's abstract describes that at least "80% of AI users" fell into the category of high scores for homework and low scores in exams. So I'd expect the distribution to skew heavily to the right.
So would I. I'm criticizing the chart, not the conclusion. Data presentation is important.
>Even if that was the case, that would be because of sampling bias
?
>the graph would still highlight a gap between in-class and out of class performance
Yes, but not a gap between AI users and non-AI users, which is what the OP was getting at.

Just to be abundantly clear, I am anti-AI. I'm just saying the OP made a shit argument.

>>25615129
>They demonstrate that, with a well-designed tool and highly motivated students, AI can enhance learning efficiency.
Every harmful technology is justified by imagining the perfect user, full of virtue and self-control, who only uses it for his own betterment, and is perfectly able to delineate the boundaries of where it should and shouldn't be used, and -- most importantly -- disciplines himself to stick to those boundaries.

But that is never how it works out. The median person spends 7 hours per day on a screen. They are not trawling Google scholar and learning how to play violin, they are scrolling through mind numbing tiktoks. Screens have indisputably made us worse off, even though THEORETICALLY if USED PROPERLY they could be good.

Same with AI. Everyone -- both the students and the professors -- will use AI improperly. Their comprehension and cognition will suffer for it. The universities will suffer for it, especially as there are so many reports now of even professors just using AI to mark people's work. I wanted to get a master's degree in philosophy, but what is even the point anymore?

I work as a software engineer and have seen first hand how the people in my company have suffered cognitive and knowledge decline as AI is integrated more and more into the workflow. It used to be that you would ask someone a question during code review, why they implemented some particular piece of code in a particular way, and they would give an intelligible answer, and you could discuss as equals and figure it out. Nowadays, people barely even understand what is going on in their own code they submit, and they have to ask their Claude before giving you an answer on anything. OF COURSE it is going to be the same in schools.

>>25616844
Oh for fuck's sake, set aside your histrionics for a moment and think
In particular, consider a student as being analogous to an athlete: does one expect an athlete, by himself, to develop his own training program, to correct his own flaws in form and conditioning, to be fully aware of technological developments and their implications for the sport, and to integrate them into his own preparation?
Now, of course schools will fuck it up; but education at-scale has never fucking worked well and it never will compared to aristocratic tutoring

>>25617014
There is 0 parallel between what you call "aristocratic tutoring" and AI. You can't tell your "aristocratic tutor" to do your homework for you. You can't make them write an essay for you and pass it off as your own. There are no "histrionics" in what I am saying. It is simply a cold and rational calculation that takes human nature into account. Humans are lazy and prone to vice and temptation. They will generally choose the easiest path. Even those who have the best intentions will sometimes get tired and fall weak to the temptation to cheat, if it is made easy for them and if everyone around them is doing it. And then little by little they will become dependent upon the machine to do their thinking. I have literally watched this happen in real time at my company, where nobody now is able to do anything without Claude.

None of these technologies that rely on human self-restraint will ever do anything good for us on an intellectual & cultural level. The internet was supposed to make us smarter because of instant access to information but it never did.

>>25617630
>The internet was supposed to make us smarter because of instant access to information but it never did.
I think the sharing of information is good but the bigger issue is the sharing of bullshit became so much easier.
It's easy to spot someone telling the truth, hunt them down and remove them

You should see the coding skills of chuck programmers (they just sit on a chair and watch).
I have coworkers who would not be hired to work as juniors anymore.

>>25613253
2% sounds about right based on the posts I see on /lit/

>>25617630
>There are no "histrionics" in what I am saying
Your hyperbolic retardation says otherwise:
>Everyone -- both the students and the professors -- will use AI improperly. Their comprehension and cognition will suffer for it
>None of these technologies that rely on human self-restraint will ever do anything good for us on an intellectual & cultural level. The internet was supposed to make us smarter because of instant access to information but it never did.
and your insistence on completely ignoring the scope of my comment to again rail about the incompetent masses makes you not worth attempting to enlighten
utterly astounding to encounter such a severe case of anosognosia

>>25617722
I am not railing against the incompetent masses. I am taking a cold, unemotional look at human psychology and predicting the future. Of course you are correct that AI can be used for learning, if used properly. But the temptation to see it as an oracle to which you can outsource your thinking is just too great. Learning requires effort, effort is painful, I'd rather relax, let me outsource this to AI. That's simply what will happen.

Homework and ai inverse proportion?

>>25612930
What field is the PhD?

>>25617698
/lit/ is considerably dumber than even the general population.

>>25617722
NTA but thus far your only arguments against him have been ad hominem and imagining a hypothetical perfect usage scenario which he already addressed.

there should be a ruling class made up of high iq individuals

>>25619444
>High iq individuals
Rules you out

File: 1789455456744249.jpg (8 KB, 320x320)

8 KB JPG

>Autist-kun reads The Diamond Age
>Let's assume AI will become the perfect tutor like in The Diamond Age - if I can imagine it, it must be possible!
>Therefore, AI WILL be the perfect tutor in 2 more weeks QED
Ashes to ashes, and slop to slop.

>>25612920
I don't understand how nobody is going ted k on these data centers. Nobody gives a shit that capital is taking everything from you

>>25613713
There's also an article by the authors that seems to cover the gist of it: https://cepr.org/voxeu/columns/generative-ai-learning-penalty-secondary-school

>>25616424
>they know how to choose the better technology, how to write superior prompts, how to validate the output better, and so on.
The entire point is that none of those skills translate to exams that don't allow use of AI. In that case, what matters is how well you've learned the actual subject matter, and learning through homework is greatly diminished if you just let AI do it for you.

>>25612930
Some of the dumbest people I've worked with at my job have had PhDs. Some are also pretty smart. "Pile it High and Deep" indeed.

>>25612920
Isn't the obvious course of action to just lower the weighting of homework to nearly 0% and raise the weighting of tests/exams to extreme highs making it so if you can't pass the final test you can't pass the class?

>>25620044
Sure but what happens to humanities disciplines? Are you really going to make philosophy students sit in exam rooms and handwrite essays out? For what purpose? Clearly, giving students the time to think, research, articulate their arguments, and spend a long time on writing their essays makes more sense. But now that's impossible with AI, so we have to impose all these artificial restrictions.

>>25612920
Schools have two populations of people. those who are motivated to study and learn and those who go to school just for the credentials. Those two populations are what these studies are uncovering

>>25616432
--Fuck off, Cunt.

If future engineers and doctors are cheating their way through exams with ai shouldn't a sensible system implement an actual competency check?

>>25612921
>BIRRIONS MUST DIE

>>25619444
so no libs and no fascists, got it

>>25619444
>there should be a ruling class made up of high iq individuals
The challenge is more how this class is maintained. They'll always exist but should it just be the best and brightest of the population representing their subgroups or a hereditary genetically distinct class?

>>25620134
>Are you really going to make philosophy students sit in exam rooms and handwrite essays out?
Is this some impossible task? When I was a history undergraduate in the early 2010s I had to do several three-hour exams which essentially amounted to writing an essay, in addition to the dissertation obviously. When I was a master's student it was basically the same, but with fewer and harder examinations. My PhD was admittedly mainly based on a thesis, but I did have to do a 2.5 hour oral viva voce examination. I gather that these vary in difficulty, mine was pretty hostile but I don't hold that against them, I wouldn't have wanted a whitewash. If I ever wanted to become a professor (in the UK a professor is a more prestigious position than lecturer) I would have to go and physically meet a panel of current professors and be tested by them to make sure I'm as brilliant as I say I am.

What is wrong with any of this?

>>25623483
Because it used to be that you could give students time to research and write the best possible piece of work they can and now they have to sit in a room and produce necessarily subpar work in totally artificial conditions, just to make sure theyre not cheating. What history book has been written in 3 hours without access to sources?

>>25623512
What are we teaching students to be good at that will help with their careers?

>>25623518
Such a stereotypically English middle class post. Next you'll be talking about health and safety.

The purpose of university education is not to "help with your career" (the very notion should make you spit). It is to edify the mind and soul, refine the manners, improve the character, and hopefully allow people to contribute to the pursuit of Truth.

>>25623518
Education should not be synonymous with job training.

>>25623537
>>25623527
I'm not concerned by specific task compliance I mean the wider philosophical foundation of how they will approach their life and career? These llm tools are here to stay in some capacity or other, The droves there purely for career padding will flounder in the face of the age of hyper distraction.

>>25614681
>>25615379
Kids who cheat are so spoiled and corrupt. It does not translate into real world success. You need to know your stuff to truly succeed.

>>25615471
They arent forced to go to school. They can just have kids. Actual teen pregnancies are people who then have to rely on their family to survive, kids can't study and work for these responsibilities.

>>25623549
>These llm tools are here to stay in some capacity or other
Data center addresses are publicly available

>>25623512
Perhaps open book tests relying on a specific textbook could work around this. You can have some access to referential materials but limited in scope, so the student is compelled to actively synthesize and retrieve relevant information from the boundary of the source and think of how to incorporate in their argument rather than simply accessing it on demand or regurgitating a summary they prompted specifically designed to answer that question instead of them coming to an answer on their own.

>>25623567
everyone has a degree now but are women better off for that? Are degrees worth less?

>>25623562
The POTUS disproves your thesis the only imperative is power you can cheat lie steal as long as you get to the top where no one can stop you

>>25624119
The leftist paradigm is a desire for power with no accountability to enforce communism

>>25613253
And what value is there in these skills for a proletariat?
We have an aristocracy on similar standing in the context of literacy.

This information should be a good thing, it is part of a growing well of evidence with which we can argue our lazy thoughts. That is, the persuasive thoughts which goad us to ignore developmental practices and oh right not everyone is schizophrenic.

>>25620134
nta but this seems perfectly reasonable. basically every inperson class in college I take now has handwritten exams, some have even gone to oral exams. colleges are basically doing what that anon said and removing homework and emphasizing in-person competence. for oral exams for instance you usually get prep time

>>25624491
That's just the objective end of all political power it doesn't matter what side you're on liberalism is a failed attempt to stop the iron law of absolutism by segregating powers between branches of government to no one's satisfaction

>>25625192
pertly true at least but one group is open about their desire for power while the other wants to achieve it while seeming like they don't.

My children will be literate. Though I can not assure it, I hope they are pedantic as well.

>>25615494
I can’t speak to historical topics like this but I’ve been using the AI as a tutor (via Claude’s built-in “learn” skill) for mathematics and it’s worked for me. In particular, I’ve found that if you include sufficient context about the key texts & textbook sections pertaining to your topic of choice then the output remains on-task and reliable for quite some time. You may have to wrangle it a bit harder to only provide responses when it’s able to find direct, primary sources supporting a given claim for studying history. Just dumping your textbook/paper .pdfs into a folder with a decent AGENTS.md should get you most of the way there tbdesu. Also remember to start new sessions on a fairly regular basis: your context window becomes functionally fucked long before whatever harness/chat client you’re using tells you it’s “full”. They’re still sycophantic, untrustworthy robots but if you get the hang of how to use it effectively for your use case it’s hard to imagine going back to the “old-fashioned way”. Also, obvs you should never be using these things to research or work on something you don’t already have a decent level of background knowledge in but it sounds like you clearly already have that.