Using AI for just 10 minutes erodes your ability to persist at hard things
Did Claude give you erectile dysfunction?
I was going to argue against this, but i don’t feel like doing the research myself.
This article is a perfect example of how legitimate research gets turned into sensationalist nonsense, and the post on Lemmy is making it worse.
The study involved 1,222 participants across three experiments. Researchers had people use AI to help with math and reading comprehension, then tested them when that assistance was removed. They found that participants were more likely to struggle with accuracy and persistence when they had to work independently.
That’s an interesting finding. It’s worth discussing.
But this whole fixation on “just 10 minutes” is ridiculous. Ten minutes of AI use under specific experimental conditions does not mean that ten minutes of ChatGPT permanently damages your brain or destroys your ability to think for yourself. The study did not demonstrate that.
There’s a pretty significant difference between showing that people can become reliant on assistance and struggle when that assistance is suddenly taken away, and claiming that AI use fundamentally erodes your cognitive abilities. One is an experimental finding. The other is an exaggeration of what that finding actually means.
And this is where I have a problem with the article and the way it’s being presented on Lemmy. The actual research is interesting enough without having to dress it up in sensationalist headlines. If the point is that people need to struggle with difficult problems to develop their skills, then make that argument. It’s a perfectly reasonable one.
But “using AI for 10 minutes erodes your ability to persist at hard things” is a hell of a lot more dramatic than what the research actually establishes.
There’s a difference between reporting scientific findings and manufacturing a narrative around them. We should be capable of discussing the potential downsides of AI without resorting to clickbait and misrepresenting what the research actually says.
Forgive me, but i’m not understanding why you find the headline to sensationalize or over state the actual findings of the study. The abstract of the paper directly says:
we provide causal evidence for two key consequences of AI assistance: reduced persistence and impairment of unassisted performance.
I’m struggling to discern a meaningful difference between “erodes your ability to persist at hard things” and “reduced persistence of unassisted performance”. Forgive me for not understanding but I would appreciate your explanation. The paper’s very own conclusion actually uses the term erosion:
Here, we show that AI interaction can result in significant impairments in independent performance and persistence – capacities that are foundational to life-long learning. If brief exposure produces measurable erosion, the cumulative effects of daily AI use over months or years may be profound and difficult to reverse
I found a definition of erosion: “the gradual loss of something as a result of an ongoing process”. Trust can erode and be rebuilt; soil can erode and be restored.
I agree that 10 mins of foundation model assistance likely doesn’t “permanently damage your brain or destroy your ability to think for yourself”. However, in my reading of the Berkeley news piece, I find no claim of this specific kind. Does the text of the press release explicitly claim the duration of effect is long-term or lasting? Does it explicitly claim the the erosion is permanent? Again, I find myself confused what you are arguing against.
“Using a compiler for ten minutes erodes your ability to write raw machine code”
This is… Unfortunately is completely true. ☹️
“Students who cheat on their homework don’t learn as much” is a better headline.
TLDR: LLMs are better information aggregators than Google, which has become a mess of ads and SEO spam. Using AI to complete homework isn’t inherently cheating; homework is practice, not an exam. If schools want to measure independent knowledge, they should use supervised assessments instead of take-home assignments. And if an assignment offers little educational value, perhaps the assignment is the problem, not the tool students use to complete it.
I think the more important takeaway from this study is that people produce better results when querying LLMs than when using traditional search engines. To me, that points to a problem with conventional search more than it demonstrates a detriment to using LLMs.
Have you tried Googling anything in recent years? It’s a bloody mess. Google increasingly prioritizes what it wants you to see, and more importantly, what advertisers have paid it to show you. Finding genuinely useful information has become an exercise in digging through sponsored results, SEO spam, and pages that exist primarily to generate ad revenue.
LLMs, for all their faults, are arguably better information aggregators at this point. That isn’t necessarily a ringing endorsement of AI, considering that they aren’t particularly good at searching either. It just illustrates how bad conventional search has become.
I wouldn’t interpret using an LLM to complete a task as inherently equivalent to students cheating. I’d argue that LLMs are simply another tool for finding, processing, and synthesizing information. If students can produce better results using those tools, perhaps we should be asking why we’re so determined to prevent them from doing so rather than examining the tools themselves.
And frankly, you can’t really cheat on homework. It’s homework, not a supervised examination. Students should be able to complete assignments using whatever resources are available to them. If the objective is to determine what students can accomplish independently, then schools should assess that through supervised examinations rather than take-home assignments.
More importantly, there’s a broader question about whether we’re assigning too much homework in the first place. Research has repeatedly found that its educational benefits depend on factors such as age, workload, and assignment quality. If an assignment provides little educational value beyond producing a completed worksheet, then perhaps the problem isn’t that students are using AI to complete it. Perhaps the assignment itself isn’t worth doing.
Have LLMs become better than Google?
Or has Google just gotten shittier than LLMs?
My SearXNG instance never lies to me, and returns results better than Google used to before they fully enshitified
Seems like a biased study yea. Why not get someone to do something manually first without AI, then give them AI to see if they do it better or worse?
Agreed, I tend to get more interested in topics and learn more than I expected, for new neural networks. I think it’s like any tool and using it right can be useful. Our brains can make new connections learning about different topics and how they are related, systems thinking and problem solving. We can search online and read the sources directly too, and verify more with increasing importance of the topic. For example, going to defend yourself in court, do a lot of research from quality sources and still reconsider lol. Just like large claims require large evidence, same idea for other resources including llms, and no need to avoid it like it’s some new Witch panic. The energy people use daily mindlessly absorbing video streams is more than text based llm use and follow up confirmations.
Learning new things and recognizing connections between topics is beneficial for the brain, as it strengthens neural pathways and enhances understanding. This process, known as neuroplasticity, allows the brain to adapt and improve its ability to learn over time
This is a really smart take on things.
What a load of shite.
If I asked an intern to do something and then review their work is that eroding my ability to work?
My managers seeming plenty eroded rn 😅
The actual study does not make the claim this stupid “article” does.
Using a lighter for 10 minutes erodes your ability to persist at starting a fire with two sticks to :)
Truth is in the middle somewhere. I can still fix/make the things I made 2 years ago, but i’m going to use AI tools to do things that it’s passable at while the tokens cost me $20 a month.
My homelab went from bare minimum with not enough time to ansible jobs to deploy any product I want to play with in a week.
Agree. With homelab being a hobby but time being a scarce resource; with AI I’m able to get past the things I hate that are mundane and that I would never find time to improve while still doing the things I enjoy with the hobby.
Also working in a job where there were never enough resources to do the tasks that were high priority, meant we also never improved on technical debt. Now that has turned: we are able to use AI assistance to deliver a higher throughput while fixing more tech debt. Just a different take than the “AI produces low quality code” position.
Yeah. I don’t reall like using AI but it has helped me a couple times with homelab and other PC related issues when even after 45+ minutes of searching the web I found nothing useful and barely anything pertaining to my issue at all.
We all knew this was going to be the case. You look at the people around you feeding their entire lives to it, surrendering their will to it completely, you knew it was going to fry their brains extra fast.
Jokes on you. I never persisted at doing hard things because I had no idea what I was doing. Now I get into all sorts of shenanigans because I do what ai tells me to do.
Shenanigans.
That’s not 37 pieces of flare!
Pretty much my situation. My local AI has whipped up some compatibility enhancements for a 2006 game, and about four save editors. Right now, I got it researching The Oregon Trail 2021’s character sheets for editing.
When a Qwen4 Flash is released, I will be looking in recompiling some Windows 3.1 games.
LLMs are a competency chipper and credibility solvent.
My observation from having to deal with entities which have engaged with LLMs, is any competency or credibility they may have had, is eroded proportionally to their level of exposure. It’s analogous to experiencing a traumatic brain injury.
Losing cognitive ability is bad enough, but it’s devastating when they never possessed it to begin with.
Yeah, if i would employ my local LLM in a “solve every simple thing for me” way, it would degrade my mental acuity too. I only defer it to areas where i don’t have any fucking clue about anything anyways or struggle very much to reach an usable result, and i do not just accept whatever was spit out, i engage with it and try to understand what is happening and why, including experimentation with my own modifications of whatever i was given.
Good examples of that is my constant battle against RegEx, sed, awk and the like for string manipulation in linux. I struggle with the syntax a lot (the rules themselves are becoming clearer to me by now, mainly because i analyze what the LLM output really does).
I started at a point where i could just confidently say that a ? is a wildcard and a * is a lot of wildcards in a row, and escaping things is something people do. Nowadays i can form most simple RegEx without consulting the LLM at all.
So we really have to put this in another framework: Do you use it to replace your own thought process? or do you use it because you genuinely do not know how to solve it and are willing to learn?
I read a study that people used to focus on remembering information they learned, but given the opportunity (Internet, books) instead they focused on remembering the location of the information they learned so that they could access it later
I don’t agree with this study and I am not a fan of AI. Replace the AI tool that they used with a calculator. You’d get the same results. Here you are legally cheating on some task because you were told to do it - breezing right along thinking of something cool you are gonna do later and then … WHAM ! Cheaty tool gets yoinked. Now fuck - I have to think ? I’d probably think fuck it, I’ll give it a half asssed shot because fuck them for making it not fun.
Replace the AI tool with a calculator. You’d get the same results.
Correct. I used to be able to calculate square roots in my head, then they let us use calculators the next school year, and soon I was no longer able to do so.
The problem is that what calculators due to math LLMs do to thinking in general, causing much deeper and extensive damage to the brain.
Also, calculators are good at maths. They produce correct results.
LLMs aren’t good at thinking (or even capable of doing it at all), they’re very good at putting the most statistically likely (given the texts they were trained on) word in front of a sequence of words. They don’t necessarily produce correct results, they produce results that wouldn’t look out of place in their training set (unless you’re an expert in whatever the text is about, in which case it’s often absolute bullshit that sounds correct to a layman).
We’re replacing working brains with autocomplete engines that (if they were capable of caring about anything, which they’re not) couldn’t care less about the correctness of their output, just that it looks correct.
And it remains to be seen if the people injured by this will ever be able to recover anything close to normal brain function. Personally, I doubt it. Once it’s gone, it’s gone.
Anecdotally, I have software I wrote entirely by hand, started by hand and ended with ai, and AI from the start. Anything I had ai touch in the last 6 months has caused this for me. I barely have the attention span to read what the agent is planning to do. I can only conceptually review at a high level and direct it that way. It gets worse every week.
I think part of it is me not spending enough time setting things up so it writes code like I would, so I am reading a strangers code really. But even then, I find myself reading it for no more than 10 seconds and being like “ugh just ask the AI”.
I have caught myself a few times about to ask the AI to change something that I could easily do myself still.
I know that if I stopped using it for a week and sat down with the code I had generated, I would understand it, but the thought of doing so has somehow become exhausting. It’s a real issue.
Yea, if you compare the post-AI test group data with the initial control group data before they had practiced, the AI test group literally did about the same as the non AI group. And you are very correct about the fact that yoinking the easy solving tool makes skipping jump because why would you do a thing that feels pointless and is harder then it was before? I hate the enviornmental and technofascist implications of AI, but people need to stop trying to make brain degradation from AI a thing, it is never going to be a thing. You just don’t really learn from it if you don’t want to y’all, like everything else. It’s not that fucking complicated. This is just a technology witch hunt.
I am begging people to look at the ACTUAL studies and analyze the variables and results critically.
I think the bigger problem is a potential for a loss of morale. We’re getting to the point where engineers can do more with less effort but it’s also less rewarding. I haven’t been this sad about doing my job in 8 years.
While I think studies like this are bunk, I think it’s a mistake to assume that there is no real effect on cognitive health.
We know that using LLMs is addictive. I can’t imagine that long term use is going to be as benign as long term use of a calculator.
It’s not going to be benign because of how much the AI can do. I caught someone using a calculator to figure out what 15% of 7000 was the other day. They haven’t calculated anything in their head in a decade so simple mathematics have become a chore.
Add that concept to basically 50% of tasks we do on a computer and the future starts looking grim.
AI is basically the devil whispering in your ear that you can be as lazy as you want, I for one find it harder and harder to resist. Told myself I’d code by hand a few hours a week to keep up my skills but that went about as far as going to the gym every week. My monkey brain loves taking the easiest path.
This is what they say about people not learning assembly and jumping straight into C++ or gasp Python
There’s a huge difference between someone using agentic ai to solve real problems quickly and someone asking a chatbot for life advice. A lot of people are asking llms for life advice, and don’t get me wrong it’s a great research tool IF you already posses critical thinking skills and know when to question it, but schools are introducing generative ai to middle schoolers before they even gain any critical thinking skills.
I agree, my main issue is people treating AI as some pseudo psychological evil. It’s just an answer box. People who have issues from it already needed help. If AI didn’t make them worse, a forum site would’ve.
It really depends on the usage. Especially people who are already mentally challenged in their normal life will probably just mindlessly use LLMs to get results, while neither understanding the task nor the solution. Had they engaged with the task the natural way, you will see in a longer series of tasks that their ability to solve these will increase, while the people just dumping stuff into chatgpt will learn nothing.
Yes, that is what I also said. The AI is not the issue, people not caring enough to do real practice is the issue.
It’s not even the caring part. Especially in the US it is the economical pressure that robs people of the time they need to make informed decisions by consulting varying sources. So i would say it’s not the caring part, it’s the part where people are pressured into a high-stress environment where even a single full-time job might not pay the bills.
Yea, that’s fair. And THAT is the thing that we should be pushing and talking about, not acting like AI is brain damage.
the calculator doesn’t get 1+1 wrong…
AI can and will be wrong and then tell you it’s right and that youre wrong and stupid for questioning it. you
AI doesn’t get 1 + 1 wrong anymore… but yes it can still be wrong. Most commonly if you ask it the wrong way. A calculator can do the same if you forget the parentheses, or mess them up slightly.
My feel is that the degree to which a wrongly inputed question can get a wrong answer is proportional to the complexity of the question you are trying to get an answer to. Kinda like computers made it possible for a simple mispelling of someone’s name to have much more massive consequences than before computers.AI does get things wrong, but it owns up to it when pointed out, and often self corrects once it has observed that something is wrong with the work. Your characterization is straight-up wrong.
"The bcrypt hooks caught only GOG web traffic (blob_007 is literally plaintext GET /token?.. HTTP) — the save crypto never touched bcrypt.dll, confirming it’s done inside the exe by statically-linked OpenSSL. That also means the ChaCha state I scanned is likely TLS traffic, not the save. Two parallel paths now: brute-force the record layout against AES-GCM (OpenSSL’s default AEAD), and locate the AES tables in the exe to find the cipher code:
No AES tables at all — the game uses hardware AES-NI, so table-xrefs are out (and that also means my earlier AES-schedule scan is the right detector… except it never finished; that was the hang). Let me nail the layout facts offline and check my tooling:"
Replace the AI tool that they used with a calculator. You’d get the same results.
Then go do the study and publish it. Should be simple as shit.
The only reasonable take in this thread is getting downvoted. Lemmy is so over.
It’s not going away. It’s a tool. Learn to use it as a tool, not a replacement for your brain.
For the first experiment with 354 participants, one group was tasked with solving 15 basic fraction problems without help, while another had ChatGPT open on the side to use for assistance and to even ask for the answer. The group with AI access started off more accurate than the group without it. Then, after 12 problems, researchers took away the AI tool. Almost immediately, the people in that group stopped solving questions accurately.
A similar thing happened when they ran the experiment on a second, larger group of 667 participants. Whereas the AI users got answers wrong or gave up entirely when the AI assistance was withdrawn, the group without AI stuck with the task and were more successful at the end.
Researchers then ran a third experiment with 201 participants and an SAT reading comprehension prompt. The results echoed those from the fraction problems: When the AI tool was removed, persistence and accuracy dropped.
I wonder if this kind of test would show the same results if you compared people using calculators for math problems vs. those without. From the basic operations to doing trig and calculus, if you provide an easier way they’re going to take it, and unfortunately not going to pick up on how the answer is obtained, which an LLM will usually tell in the process. Doing something yourself always teaches you better than watching it, or just being told the answer.
If the tests are implying or showing that people who already were proficient in doing these tests and would have done well without started declining after exposure to an LLM’s help, and that persisted… that’s a different problem and one that should be looked into. Losing skills you had because of help? That shouldn’t happen.
A lot of pro AI people like to compare LLMs to calculators but it’s not comparable at all unless the math you’re doing is simple arithmetic. If you’re doing calculus, the skill you’re working on isn’t arithmetic, so the calculator isn’t much of a factor. Lots of people forget how to do long division after using calculators on the other hand.
Losing skills you had because of help? That shouldn’t happen.
That’s exactly what should happen. Cognitive thinking skills are use it or lose it. If you use an LLM to replace your brain, your brain will get worse.
Tech evangelists rely heavily on logical fallacies. False equivalence is one of the main ones. “A thing happened before therefore this tech is nothing to worry about”. They’re redirecting the argument to bait people to try to prove why calculators are bad. But of course the past already happened and nobody is going prove otherwise.
Yep, especially in mathematics a calculator is not much of a help in calculus. The moment you put numbers in whatever thing you wanna solve is the point where the hard part is already done.
Speak for yourself lol. I failed calculus the first time because my algebra skills were shit and it had nothing to do with calculus
That’s an issue with your teacher putting the focus on the wrong part. Sorry you had to experience that. My math skills (which were mediocre at best at the time i left school) skyrocketed the moment i got a teacher that made sure we understood the calculus part and didn’t even want us to insert real numbers.
I hate that I never had a good math teacher and that the ones I had made me hate maths.
I would love to like maths, but there’s too much trauma there.
I’ve taught myself what I’ve needed for work, but it has never been a pleasant experience.
Bad math teachers are a plague on humanity.
but it’s not comparable at all unless the math you’re doing is simple arithmetic.
That’s what they were studying on the math side
For the first experiment with 354 participants, one group was tasked with solving 15 basic fraction problems
Also
If you’re doing calculus, the skill you’re working on isn’t arithmetic, so the calculator isn’t much of a factor.
Calculators would be a big factor in solving calculus problems. Haven’t done calculus in a while but if I recall correctly you are breaking down or rearranging the problem into smaller steps that are more basic arithmetic. One part of a calculus problem could be finding out what the square root of 7 is , if you have a calculator to do that you’re more likely to solve the problem more accurately and to not give up on the problem, which were the two metrics this study was looking at.
Realistically, you would just write root 7 as a coefficient of your answer if you’re performing this math long hand. No one really expects anyone to be able to calculate things like that.
Over time yes, but this says skills were lost during a single session or short term. It was a random group of people, so unknown on what skills they did have before. I think what they observed wasn’t a permanent loss, but a placebo effect of having extra help and it was taken away, so for the rest of the test they felt lacking.
My example of calculators was implying one that could do the test work. All calculators have trig functions, and some have calculus abilities. Would a student forget how to find cosine if the calculator was taken away, or do a derivative of a function? Not in one test. But if they were using it first half and then had to do it themselves for the rest, I imagine there’s some psychological stuff going on that makes you feel like you’ve lost that skill.
They were tested on fractions
Did you use an LLM to summarize the study for you or something?
And reading comprehension.
We’re discussing the effects of the test, not what the test was.
The content of the test is relevant to the outcome
Doing something yourself always teaches you better than watching it, or just being told the answer.
A neat result from pedagogical studies is that being given the worked solutions to new problems to study from is much more effective for learning the material than simply practicing.
Students who were assigned practice problems performed worse on later tests than those given the same problems and all of the worked-out solutions for them. The explanation from the research was - when learning a new topic, immediate practice isn’t beneficial because the student doesn’t yet know what to practice. If they are lost when trying the problem, they can easily end up practicing the wrong skills by not knowing where to start. Those with the solutions have some of the cognitive burden removed, and are better able to internalize the task and generalize it to later problems.
I completely agree with you on LLMs and I think they’ll be a disaster for education. But I’ve always found it interesting that being given the answer can be useful for learning in certain circumstances.
A neat result from pedagogical studies is that being given the worked solutions to new problems to study from is much more effective for learning the material than simply practicing.
Finally, something to justify my still-existent hatred of my middle school geometry teacher for literally telling us ‘the theorems are on the board’ any time we asked her for help trying to solve a problem or for her to show us an example problem. Why the fuck were you even teaching.
That was one of my gripes with my experience in college vs hs. In my hs at least, we were taught the most difficult combination of a problem and then everything that we practiced would be able to be worked out from there. In college, we were often taught the simplest example and somehow expected to derive the most difficult version of it on the problem sets. That’s how you get students who study really hard for an exam and half of them fail. They never taught the proper way to solve the problem!
I was out of college by the time LLMs came out but I still believe they could be helpful in this scenario if fine tuned and limited as a tutor not an answer-giver. Obviously if you have access to a good human tutor (who doesn’t call you dumb like my friend had at the time), but I didn’t have access to one.
There are different ways to learn, some work better for certain people than others.
I would think that the best way to learn is neither of these options, but to be taught how to do the work by someone who understands (like a teacher) and then made to do it after it has been thoroughly explained to you and practiced with you.
That’s what works best for me anyhow, that or literally reading the textbook, lol, but that’s less efficient time-wise.
Immediate practice on something you haven’t been walked through yet? I guess the way I phrased it made it sound like jumping in yourself with no guide is better. You definitely need to be taught first, or at least as you go, and not figure it out on your own. The point was that simply watching someone do something isn’t enough, you have to develop the process of getting to that point yourself. But only after seeing that route, otherwise you will take lots of wrong paths.
Calculators don’t help with reading comprehension (3rd Test)
Also, they’re not talking about trig and calculus, they’re talking about fractions
You know (hopefully), like 1/5 =0.2? Or 6/18=1/3=0.333?
Yes, I’m aware of what they tested and what fractions are…
I was just expanding the point to other things that are comparable. And an LLM for a reading comprehension test is exactly what a large language model would be best suited for, so in essence a language “calculator”.
Doesn’t change the rest of the discussion, which was about the brain’s ability to jump back into things when the crutch is taken away.
I don’t think it’s fair to say that fractions are similar to trigonometry and calculus aside from the fact that they’re all types of math
Very few people do trigonometry or calculus in their heads, I’m sure most people who know it really well could do a cosine on paper, but that’s a much higher level operation than a fraction
Um, you know all of those were developed before calculators, right?
I learned both trig and calc on the principles in how they were derived, so yes, when I was learning and using them, I could do some of it in my head or with a few figures on paper.
But I never said “similar” at all, I was using them as comparable subjects that could be done by hand or with aid.
You’re so obsessed with calling out someone on not mentioning fractions you’ve missed the point of the test, article, and this thread. They use fractions because it was above simple math but not so high level that some of the random people picked might not know it. Which going back to the real discussion, having randoms whose skill isn’t known even halfway through the test might have been a flaw in the test.
Um, you know all of those were developed before calculators, right?
What does the timing of the advent of the calculator have to do with what I said?
But I never said “similar” at all, I was using them as comparable subjects
I would contest that they are comparable subjects. Entry level piano lessons and music theory are both technically of the same subject, but wildly different in practice. One is suitable for beginners or lay people, the other is not.
You’re so obsessed with calling out someone on not mentioning fractions you’ve missed the point of the test, article, and this thread.
You are projecting your own obsession of using the poor comparison of “AI = Calculator” onto me. That’s why I keep bringing up fractions, it is an easy-to-understand (for most) rebuttal to your logically fallacious proposal.
They use fractions because it was above simple math but not so high level that some of the random people picked might not know it.
Exactly, that’s why your comparison doesn’t work.
Which going back to the real discussion, having randoms whose skill isn’t known even halfway through the test might have been a flaw in the test.
What changes would you make to the experiment to remove the flaw you’re perceiving?
How do you expect that change would impact the outcome?
As I’ve said before, the abilities of the ones with ChatGPT assist don’t seem to be known, as they started off with help. The assumption is that they might on average be the same as the others, but we don’t know that. A simple fix to start is to offer the LLM help in the middle of the test for them, see how the curve of ability goes up and down, and compare before and after.
Not answering the rest, I was more interested in the study and what it was claiming and not arguing about missing context of commentary. I think the study may be right in suggesting people today have more trouble understanding things than they used to, but it’s far deeper and older than LLMs. They are just the latest part of the bigger problem.
Some kind of comparison group would have made the result more informative.
Replace this with math and calculators. Similar results? Then it’s not AI’s fault, but a known phenomenon.
Of course, they would never even get the funding for this “experiment” if they didn’t have AI fearmongering in the title.
You can’t do fractions without a calculator?
I assume you’re also a fan of AI then?
A calculator doesn’t stroke your ego and you know that. That’s why you like and use AI.
As someone who has a disorder and is prone to hyper fixation, this isn’t a concern. The largest issue is that AI often claims they know how to do something, and it’ll get you about 80% the way there.
The at the 80% mark, it admits the thing you’re doing can’t be done with the software or tool it recommended.
I don’t know the details of this study but it sounds like a basic result if you get people to rely on something for most of an activity and then you take it away vs those who are doing it manually, they’ve been given more progressively difficult challenges but they’ve been actually practicing and doing the work and hence internalizing it better.
It’s like those open versus closed book exams where those who actually study can do much better than those going through a book during the exam to figure it out. It’s about internalizing and practicing a material.
Not really surprising results. I guess the empirical verification is good.
I feel like they could have (and probably did) say the same think about search engines 25 years ago.
Though this does seem a bit worse, depending on how you use it.
Yes, I was going to ask, how does this compare to with/without internet access. Coders were also likely more persistent when they could only read the tool’s manual and keep trying, vs getting the answer from StackOverflow immediately.
I’ve heard there’s been a decline in memorization skils since the internet, but I have not verified.getting the answer from StackOverflow immediately.
StackOverflow didn’t have immediate responses and the responses were how to do the thing, not literally writing the code for you. The internet and support communities were like reference books and asking a colleague for advice, the normal way people learn.
LLMs are like having someone else do it for you. You have to actively try to lean something from how it did it and most people don’t, they just take the answer and move on.
Stackoverflow definitely had code you could just copy and that would be the end of it. Not always though.
Dictionaries have definitions you can just use!