"At the other end of the distribution, AI students who spend more than 65 minutes on their homework receive homework and exam scores similar to those of non-AI students, suggesting that these students do not use generative AI for homework assignments. However, this group consists entirely of students who adopted generative AI no more than Öve months. Six months after adoption, no AI student spends more than 65 minutes completing their homework (see Figure A5). This is consistent with the gradual process of learning how to use AI tools. It also suggests that AI crowds out the highest level of e§ort."
"Interestingly, in the range of 50-65 minutes, the median and the interquartile range of exam scores of AI and non-AI students are similar. This implies that, in the range where AI students and non-AI students have overlapping homework times, students who spend the same amount of time completing homework on average receive similar exam scores."
"This pattern shows that students who spend the same amount of time on homework learn similarly, with or without generative AI. In other words, generative AI reduces time spent learning for the majority of AI students but not learning efficiency for those who spend the same time studying as the non-AI students."
I'm confident it's an amplifier for people who know how learning works and already do a lot of it, successfully. However the level of "learning fluency" I'm talking about isn't reached for many until late college or grad school, and sometimes not at all. So I'm not surprised by the quoted results for 12-18 year olds.
"The negative learning effects are larger for students with higher initial achievement. The differences in the estimated full (6-10 month average) effects are substantial, with a 50% gap between the most negative effect (-24 percent) for the highest tercile and the least negative (-16 percent) for the lowest tercile. "
Not top 10% as you asked, but the closest to what you asked. My working hypothesis is that top performance is highly correlated with willingness to work hard, and AI decreases the motivation to work hard.
they're not designed to measure general aptitude, or function as admissions criteria, or screen for job applications, or any other numerous things they are used for.
there can be many questions of pedagogy. one of them is, what do our exams measure and how do we use them? professors who say, "My exam is designed to measure who studies, not be used for all these other purposes that they are actually used for" - I don't buy it. It's the same as late night comedians saying they are not responsible for solutions, even when spending 90% of their air time making political jokes.
THIS is the pedagogical issue, that pedagogy has NEVER caught up with the scope of responsibilities. This is acute in STEM - I mean, the humanities departments are generally pretty well run, all things considered, in this regard. Generative AI is accelerating that pre-existing crisis.
Huh? They're designed to measure how much you know. They can't see how much you study, nor would they have reason to be interested.
At the end of the day though what matters is what you know. Furthermore, if it's a serious subject, it shouldn't matter whether you learned it from this teacher or from another school and teacher, as long as your knowledge is correct. Knowing the idiosyncracies of this particular teacher should not factor into the grade. A serious subject can be learned on one continent and examined on another. Bullshit courses are all about learning pet peeves and hobby horses of a particular teacher.
The "slightly higher" performance is based on statistically insignificant samples (between 4 and 20 students, depending on the context, out of the total population of 26,000): https://bsky.app/profile/benjaminjriley.bsky.social/post/3mt...
Citation needed? I have no clue where you got this from. I hadn't even heard of it as a conjecture, let alone as something anyone accepted, let alone as gene rally accepted...
1. They don't do any homework.
2. All the in-class time is split between the teacher babysitting and playing social worker to problem students, and lecturing, with little to no opportunity to actually practice what they've learned?
I understand that some students don't have home environments that are conductive to doing homework well. I understand that some students are enrolled in five hours a day of extracurricular university-application-padding activities. I understand that some students have incredibly poor screen discipline and impulse control.
But I don't understand that anyone has magically figured out how to teach complicated things to students, and have it stick without them spending a lot of time practicing what they are learning.
As anyone who has tried to do something hard knows, the first step to being good at something is to spend a lot of time being pretty shit at it.
A student who has written and received feedback on 500,000 written words is going to be way better at writing than that same student who wrote 50,000, just like someone who has put 5,000 hours of focused practice into playing the piano is going to be better than my dumb ass, who has only put 100 hours in.
(If you found the solution to get good at stuff without practicing it, I'd love to get good at piano without putting any homework in on it.)
Same can be said of technology in general tbh.
Suppose it's good to learn how elastic the brain is, in both directions, at a young age where it doesn't matter.
We need to shift the incentives by adding ruinous penalties for things that are currently quite commonplace if they are done by large players. Some dude training his own AI on his own computer can scrape and train. The fine for OpenAI or Meta using a single copyrighted book without permission should be in the tens or hundreds of millions.
What we're seeing currently in our society is a "loophole inversion" where the rules have an effect mainly via their loopholes. The most profitable activity is to find loopholes and exploit them as frenetically as possible to gain as much advantage as you can before the loophole is closed, or get people hooked on the loophole so it's retroactively legalized. Entities that are big enough to do this are big enough because they have lots of money behind them. Entities doing the same kinds of things without lots of money are not really doing much harm. So the best approach is to adopt a "sliding scale" in which even tiny violations by wealthy actors result in penalties enormously greater than fairly large violations by small players.
LLMs make significant mistakes frequently and smart people have no way of judging those mistakes outside their domain expertise. They are also sycophantic and great at being an echo chamber which makes people feel smart even if they are not.
So I think the burden of proof is on you to prove that they somehow amplify intelligence, it seems highly unlikely.
All of those sound like flaws and defects of dumb people?
Smart people know LLMs confabulate and tell them they’re Absolutely Right! Smart people don’t want to be embarrassed by trusting the hallucination machine and revealing their gullibility to others.
Raising the noise floor like this only makes it that much harder to find "Smart" people, which we were already doing terrible at.
I use Claude every single day, but this is such a bad tradeoff. Maybe it will help me standup a quick fix when that is needed. Maybe it can help me dig through documentation to find relevant bits and figure out the unstated assumptions underlying it. Maybe it helps me generate test cases.
Meanwhile, my day to day life is now noise. All social media is noise. All content is noise. Slop pours onto me from all directions. Writing more test cases isn't helping me.
Am I smart? Am I dumb? I don't care, right now I'm deafened
I'm using Claude at work myself and am impressed with the product, but notice that this is the only reason I need to use it at all. Our product pages were shit to begin with, now they're AI-generated and somehow even worse. Our procedures are incomprehensible spaghetti with enough arbitrary context switching to give a sadistic Soviet municipal administrator an erection at the thought of watching anyone try to actually follow them.
Use AI to create inefficiencies, then use AI to bypass them. Those who can't do the latter will struggle to survive.
Clarification: to value “smart” people, which we were already doing terrible at.
It does give us a new heuristic, though: people who are willing to completely cut generative AI out of their lives (cold-turkey, if you ever started using it) are a much smaller group of, predominantly thoughtful, people. You do have to give up Claude to be part of this group, but from what you say, that's no great loss, and no longer being deafened is worth it.
This has considerable advantages over conventional elitism, because the barrier-to-entry is negative in almost all cases.
The one exception I've found is assistive tech, where the state-of-the-art is so poor that vibecoded slop is genuinely an improvement over the state-of-the-art, and in many cases the tooling simply isn't available to make your own assistive tech (unless you want to bootstrap an entire networked computing environment, which isn't very helpful when you want to do your online banking and do not, in fact, work at your bank).
But there are not many principled exceptions where you could seriously argue that the trade-off is worth it. Take mathematics, for example, which we often see touted as a "good use-case" of generative AI. The primary advantage of generative AI in mathematics is being able to search though a vast corpus of ivory towers and inconsistent terminology (without proper attribution) to locate and connect ideas that can help solve problems. The deficiency this is addressing is elitism, inadequate communication, and inadequate indexing within academic mathematics. This problem is entirely created by the academic mathematicians, and has been known for nearly a century (per https://en.wikipedia.org/w/index.php?title=Nicolas_Bourbaki&...):
> Bourbaki was founded in response to the effects of the First World War which caused the death of a generation of French mathematicians; as a result, young university instructors were forced to use dated texts. While teaching at the University of Strasbourg, Henri Cartan complained to his colleague André Weil of the inadequacy of available course material, which prompted Weil to propose a meeting with others in Paris to collectively write a modern analysis textbook.
To my knowledge, this is the only organised project to clean up and improve mathematical communication. Everything else (Metamath, Mizar, AFP, Lean) is yet another ivory tower. The Wikipedia article on this topic (https://en.wikipedia.org/wiki/Mathematical_knowledge_managem...) risks deletion as non-notable, that's how little anyone's actually trying. They made their own bed, and generative AI will only provide a brief respite from having to lie in it. (I was surprised how many other "compelling" use-cases evaporated when I applied this razor to them: the sibling comment https://news.ycombinator.com/item?id=49392265 points out one such.)
Vibe-coding assistive tech which doesn't yet exist, as a temporary scaffold to improve the quality-of-life of yourself and others in a social world dominated by non-essential access barriers is, to my knowledge, the only exception to this principle that can be justified. If you treat people who make other excuses, or who don't even bother with excuses, as not worth listening to, you lose little – and doubly-so, if you make your stance clear, so that others know the "cost" of gaining your attention.