Thoughts on thinking

Completely agree.

From all of my observations, the impact of LLMs on human thought quality appears largely corrosive.

I’m very glad my kid’s school has hardcore banned them. In some class they only allow students to turn in work that was done in class, under the direct observation of the teacher. There has also been a significant increase in “on paper” work vs work done on computer.

Lest you wonder “what does this guy know anyways?”, I’ll share that I grew up in a household where both parents were professors of education.

Understanding the effectiveness of different methods of learning (my dad literally taught Science Methods) were a frequent topic. Active learning (creating things using what you’re learning about) is so much more effective than passive, reception oriented methods. I think LLMs largely are supporting the latter.

Anyone who has learned a second language can tell you that you aren't proficient just by memorizing vocabulary and grammar. Having a conversation and forming sentences on the fly just feels different- either as a different skill or using a different part of the brain.

I also don't think the nature of LLMs being a negative crutch is new knowledge per se; when I was in school, calculus class required a graphing calculator but the higher end models (TI-92 etc) that had symbolic equation solvers were also banned, for exactly the same reason. Having something that can give an answer for you fundamentally undermines the value of the exercise in the first place, and cripples your growth while you use it.

Same with drawing which is easy to teach, but hard to master because of the coordination between eyes and hand. You can trace a photograph, but that just bypass the whole point and you don’t exercise any of the knowledge.

This reminds me how back in my school days I was not allowed to use the internet to prepare research on some random topics (e g. history essay). It was the late 90s when the internet started to spread. Anyway teachers forced us to use offline libraries only.

Later in the university I was studying engineering. And we were forced to prepare all the technical drawings manually in the first year of study. Like literally with pencil and ruler. Even though computer graphics were widely used and we're de facto standard.

Personally I don't believe hardcore ban will help with any sort of thing. It won't stop the progress either. It's much better to help people learn how to use things instead of forcing them to deal with "old school" stuff only.

You can learn a lot from LLMs though, same with, say, Wikipedia. You need curiosity. You need the desire to learn. If you do not have it, then of course you will get nowhere, LLMs or no LLMs.

Another case in point is that memorizing vocabularies and grammar, although could seem like an efficient way to learn a language, is incredibly unrewarding. I've been learning japanese from scratch, using only real speech to absorb new words, without using dictionaries and anything else much. The first feeling of reward came immediately when I learned that "arigatou" means thanks (although I terribly misheard how the word sounded, but hey, at least I heard it). Then after 6 month, when I could catch and understand some simple phrases. After 6-7 years I can understand about 80% of any given speech, which is still far, but I gotta say it was a good experience.

With LLM's giving you ready-made answers I feel like it's the same. It's not as rewarding because you haven't obtained the answer yourself. Although it did feel rewarding when I was interrogating an LLM about how CSRF works and it said I asked a great question when I asked whether it only applies to forms because it seems like fetch has a different kind of browser protection.

Never underestimate laziness, or willingness to take something 80% as good for 1% of the work.

So most are not curious. So what do you do for them?

Can you learn a lot? Or do you get instant answers to every question without learning anything, as OP suggests?

You can learn a lot, if you want to. I can ask it a question with regarding to pharmacodynamics of some medication, and I can ask more and more questions, and learn. Similarly, I could pick up a book on pharmacology, but LLMs can definitely make learning easier.

You have to somehow figure out the root cause of the laziness, or if it really is laziness, and not something else, e.g. a mental health issue.

Plus, many kids fail school not because of laziness, but because of their toxic environment.

Well I can extract a square root by hand. We all had to learn it and got tested on it.

No one to day learns that anymore. The vast, vast majority have no idea and I don’t think people are dumber because of it.

That is to say, I think it’s not cut-and-dried. I agree you need to learn something, but something’s it’s okay use a tool.

> I’m very glad my kid’s school has hardcore banned them.

What does that mean, I’m curious?

The schools and university I grew up in had a “single-sanction honor code” which meant if you were caught lying or cheating even once you would be expelled. And you signed the honor code at the top of every test.

My more progressive friends at other schools who didn’t have an honor code happily poo-pooed it as a repugnantly harsh old fashioned standard. But I don’t see a better way today of enforcing “don’t use AI” in schools, than it.

I was expecting some response like this, because schools have “banned” things in the past.

While this is superficially similar, I believe we are talking about substantially different things.

Learning (the goal) is a process. In the case of an assignment, the resulting answer / work product, while it is what is requested, is critically not the goal. However, it is what is evaluated, so many confuse it with the goal (“I want to get a good grade”)

Anything which bypasses the process makes the goal (learning) less likely to be achieved.

So, I think it is fine to use a calculator to accelerate your use of operations you have already learned and understand.

However, I don’t think you should give 3rd graders calculators that just give them the answer to a multiplication or division when they are learning how those things work in the first place.

Similarly, I think it’s fine to do research using the internet to read sources you use to create your own work.

Meanwhile, I don’t think it’s fine to do research using the internet to find a site where you can buy a paper you can submit as your own work.

Right now, LLMs can be used to bypass a great deal of process, which is why I support them not being used.

It’s possible, maybe even likely that we’ll end up with a “supervised learning by AI” approach where the assignment is replaced by “proof of process”, a record of how the student explored the topic interactively. I could see that working if done right.

I don’t see the problem.

I’m not sure how LLMs output is indistinguishable from Wikipedia or World Book.

Maybe? and if the question is “did the student actually write this?” (which is different than “do they understand it?” there are lots of different ways to assess if a given student understands the material…that don’t involve submitting typed text but still involve communicating clearly.

If we allow LLMs- like we allow calculators, just how poor LLMs are will become far more obvious.

If LLMs are allowed then sure. However, when LLMs are explicitly banned from use, is the case I am talking about.

You can learn an incredible amount. I do quite a bit of research as a core part of my job, and LLMs are amazing at helping me find relevant research to help me explore ideas. Something like "I'm thinking of X. Does this make sense and do you know of any similar research?" I also mentor some students whose educational journey has been fundamentally changed by them.

Like any other tool, it's more a question of how they're used. For example, I've seen incredible results for students who use ChatGPT to interrogate ideas as they synthesize them. So, for example, "I'm reading this passage PASSAGE and I'm confused about phrase X. The core idea seems similar to Y, which I am familiar with. if I had to explain X, I'd put it like this ATTEMPT Can you help me understand what I'm missing?"

The results are very impressive. I'd encourage you to try it out if you haven't.

> if it really is laziness, and not something else, e.g. a mental health issue.

Kids optimize. When I was in high school I was fully capable of getting straight F's in a class I didn't care about and straight A's in a class I enjoyed.

Why bother learning chemistry when you could instead spend that time coding cool plugins and websites in PHP that thousands of internet strangers are using? I really did build one of the most popular phpBB plugins and knew I was gonna be a software engineer. Not that my chemistry professor cared about any of that or even understood what I'm talking about.

I've used it these past few months to better understand the PDF format, Nix, and a few other technical concepts.

I try to use AI to automate things I already know and force myself to learn things I don't know.

It takes discipline/curiosity but it can be a net positive.

How much hours would you estimate did you watch (I assume it was video, not just audio) in those years? What kind of material? Just curious.

From the article:

“The irony is that I now know more than I ever would have before AI. But I feel slightly dumber. A bit more dull. LLMs give me finished thoughts, polished and convincing, but none of the intellectual growth that comes from developing them myself. The output from AI answers questions. It teaches me facts. But it doesn’t really help me know anything new.”

I think the thesis is that with AI there is less need and incentive to “put the work in” instead of just consuming what the AI outputs, and that in consequence we do the needed work less and atrophy.

I very much agree with your sentiment here.

I tried to encapsulate that to some degree when writing something (perhaps poorly?) recently actually - https://smcleod.net/2025/03/the-democratisation-paradox-what...

Comparing extracting a square root my hand is rather different in scope than reducing / simplifying equations entirely. The TI-92 could basically do all of your coursework for you up to college level, if memory serves.

The real question isn't "is it okay to use a tool" but "how does using a tool affect what you learn".

In the cases of both LLMs and symbolic solving calculators, I believe the answer is "highly detrimental".

The school has an academic honestly policy which explicitly bans it, under “Cheating”, which includes:

“Falsifying or inventing any academic work, including the use of AI (ChatGPT, etc)”

Additionally, as mentioned, the school is taking actions to change how work is done to ensure students are actually doing their own work - such as requiring written assignments be completed during class time, or giving homework on physical paper that is to be marked up by hand and returned.

Apparently this is the first year they have been doing this, as last year they had significant problems with submitted work not being authored by students.

This is in an extremely competitive Bay Area school, so there can be a lot of pressure from parents on students to make top grades, and sometimes that has negative side effects.

Feels different, comes naturally, without conscious thought, just like we don't focus on beating our hearts.

And agree about learning by practicing a skill being best. But you and I both know the school system has worked on rote memorisation for hundreds of years at least and still is now.

Honestly, I doubt that LLMs are great for learning. Too often, they output plausible-sounding things that turn out to be completely wrong. I know Wikipedia can have its problems with factuality, but this is on an entirely different level. (And yes, they do this even when they're allowed to do web searches and "reason".)

The effort of verifying everything it claims may or may not outweigh the effort of other means of learning.

> I’m very glad my kid’s school has hardcore banned them.

Schools will ban anything they think of as sinister.

Huh? While I essentially never have need to compute a square root by hand (unless it is a perfect square of course), shouldn’t one know how one would?

Ironically, states now use AI to grade student essays in standardized tests.

English teachers even recommend grammarly..

Students are given a “prompt” for writing.

I wish other schools had the conviction you describe…

The manual methods are also the foundation for higher approaches involving approximation and iterative solutions. These are widely used in engineering and science.

Pressing a calculator key doesn’t give the same insight.

Today such infractions might result in a verbal warning…

Oral presentation without notes and a live Q&A would be some ways…

Realistically, putting them into trades sooner could almost be a good thing. Kids who don’t want to learn end up dragging down the class and distracting those who do.

But, these are kids… Hard to argue that adults should selectively deny education when it is their responsibility to do otherwise.

We don’t neglect the handicapped because it is inconvenient to provide them with assistance.

Using a tool like that is opposite to mastering the skill. There's no royal road to mastery and never will be. One does not have to master all skills, of course, and may do well not mastering any (or mastering dark ones).

Wikipedia isn't going to write your paper for you. I don't see the difference between an LLM and one of those paper writing services in this context.

> Students are given a “prompt” for writing

Students were always given a “prompt” for writing.

That’s why tech companies used that term; rather than the other way around.

> the higher end models (TI-92 etc) that had symbolic equation solvers were also banned

I'm surprised it was a problem in the first place. Don't equation solving exercises require you to leave intermediary steps, and you can't just put a "x=5" as a one liner answer ?

Yeah and I'm of the age when teachers in all grades would say "you're not going to carry around a calculator your whole adult life"

Hilarious miscalculation.

Why should one? Perhaps they should if it's relevant to their work, daily routine, or interests. But if they have no need for it?

> Students are given a “prompt” for writing.

What do you think "prompt" mean?

Or you're saying the students are asked to mimic AI's style?

Yes they are dumber because of it, not in the mental retardation kind of way but a more nuanced way, among others the mental work you put into trying to find another simpler way than the one the professor is teaching you, and the understanding about numbers such attempts can give you, even if they are unsuccessful.

I am waiting for the day (maybe it's already here) when I can talk to an LLM to practice my 2nd language. It can correct everything I say, it can talk forever, it can challenge me to use new grammar or vocabularly. Note: I can speak all day in my 2nd language with friends but I wouldn't give a business presentation nor could I explain, as a native, how something technical works. If I watch a TV show I might understand 30%-99% but the more lawyers/military/goverment/science parts there are the more it's beyond my current level.

Getting exposure there is hard. Talking to friends just means more practice with what I already know but an LLM could help me practice things outside that area.

Needs are all fabricated, Ludwig Wittgenstein said "the limits of my language are the limits of my world", the same thing happens with logical thinking and all its tools including math.

For many languages, this is already something you can do.

Asking as a non-American non-school-pupil-parent: what does it mean for a school to be competitive in this context? Competitive entry into a school I understand, but that threshold has been cleared. Isn’t US college admission based on essays and standardised tests like GMAT, SAT, GRE?

I know, that is why you need the desire, the will to learn. I have been using LLMs for this, so I know it is possible. I understand what you are saying though, and it is indeed a sad state of affairs, but then again, this was the case due to search engines, Wikipedia, and so forth, long before LLMs.

Again, you can truly learn a lot using LLMs, but you have to approach it properly. It does not have to be just "facts", and sometimes, even learning "facts" is learning.

I can use LLMs and learn nothing, but I can use LLMs to learn, too!

Thank you, and the previous commenter. I am tired of trying to convince people that LLM can be a really good tool for learning. :/

They should just simply try it. Start with something you actually know to see how useful it might be to you with your prompts.

We are talking about learning. You can learn much more from LLMs than Wikipedia, because if you do not understand something, you can always ask an LLM about it, and it would reply to you in any way you want; whatever helps you learn better.

Yeah, I remember reading someone saying you won't use a fork lift in a gym. I think this is the same idea.

The problem is really about how to evaluate performance or incentivize students to actually work on their exercise.

I don't remember if it was the case for the TI-92+, but some calculators can show the intermediate steps, or at least some of them.

> states now use AI to grade student essays in standardized tests.

citation needed

If you used subtitles over audio then why would you avoid dictionaries too ? Purely for the reward of treating it as a puzzle ? (Since you would have to figure out which word corresponds to a which concept in a phrase.)

What you just described is irrelevant to what we are discussing.

As for what you said, yeah, I got 1s (Fs) because I was too busy coding and reading books on philosophy, as a 14 years old.

> No one to day learns that anymore. The vast, vast majority have no idea and I don’t think people are dumber because of it.

Arguably, the kind of person who was helped by learning to do that by hand still learns to do it by hand, but because of curiosity rather than because a teacher told them to.

I remember being thirteen and trying to brute force methods for computing the square root. I didn’t have the tools yet to figure out how to do it in any systematic way, and the internet wasn’t at a point yet where it would have even occurred to me to just search online. Wikipedia wouldn’t exist for another two years.

I probably finally looked it up at some point in high school. I’m not sure exactly when, but I remember spending a lot of time practicing doing a few iterations in my head as a parlor trick (not that I ever had the opportunity to show it off).

If I were thirteen and curious about that now, I’d probably just ask ChatGPT. Then I’d have a whole follow up conversation about how it was derived. It would spit a lot of intimidating LaTeX at me, but unlike with Wikipedia, I’d be able to ask it to explain what those things meant.

This is the thing I don’t get when people talk about LLMs’ impact on education. Everybody focuses on cheating, like learning is inherently a chore that all students hate and must be carefully herded into doing despite themselves.

But that’s a problem with school, not learning. If your actual, self-motivated goal is to learn something, LLM’s are an incredible tool, not a hindrance.

Mostly anime. Surprisingly, not that much, I think somewhere in the ballpark of 100 titles. In the beginning I was also watching some grammar tutorials on YouTube to get started with grammar quicker (Otherwise convergence on solution would be too slow).

Contrary to what I said I actually did use dictionaries, but the point I was trying to make is rather than memorizing phrases in advance, I used it to translate something I thought I heard.

Leave ‘em behind and win the race.

That's an surprisingly "strict" (in quotes for obvious reason) honor code.

I'm at some uni in Poland, not top tier, but at the same time - not bad either, slighly above average.

The amount of cheating I saw - it's almost mundane. Teachers know this, so do we...

Physical paper isn't going to save them.

(Also, typing was only appropriate for only some classes anyway.)

Yes, but previously you didn’t need the desire that much, because you were more forced to it, there was no easy way. The fact that now you need that internal motivation means that it will happen less, where previously it happened by default.

How is it irrelevant? Kids will always cheat their way through classes they feel are a distraction. Even the super smart Type A kids.

Hell, all humans do that. You use every resource available to get out of dealing with things that are not your priority. This means you will never be good at those things and that’s fine. You can’t be good at everything.

I agree, it is sort of like a double-edged sword, I would say.

They will, but we were talking about the will or motivation to learn. If someone has a curious mind, and actually wants to learn, then they can definitely use LLMs to do that.

I don't disagree with you though.

you put my thoughts into words I couldn't.

Any school's #1 job is to motivate learning. Schools clearly suck at this.

LLMs are a fascinating effective learning tool. early learning would be better off embracing it as such.

i imagine a combination of a video of a good tutor explaining a concept followed up with an llm to quiz and explain the concept seems far better than what we have today.

that's on them? Why do I have to take responsibility for someone else's growth?

I think we need to find a way to teach without grades. That is, we do know how to teach without grades at younger levels, and many schools do it successfully.

The problem is that eventually you need to measure for placement, and Goodhart’s Law kicks in and destroys the enjoyment of learning. It’s very hard to be intrinsically motivated when the external pressure dominates.

The fact that “study for the test” is basically synonymous with “study” for most people is indicative that we’re doing something extremely wrong in education.

Memorization and manual methods also help develop intuition about what is a plausible answer and what is not. It helps build a brain that is able to sanity-check what it is being told.

It worked. There was still cheating (caught and uncaught), but 50-100x less than what I saw at other schools.

And it gave students a sense of pride in their education

This was already happening a decade ago lol. Happened to me at least.

> The problem is that eventually you need to measure for placement Why do you need to be "placed"? I know the answer - resources.

But why can't each individual have sort of their own bespoke governess tutoring them all through school? Pay Gilbert Strang a million dollars to do a 32 week course in LinAlg. Yes I know a normal course is shorter, expand it to cover everything in exquisite detail, so even I can keep up. Alan Kay, Donald Knuth, and John Carmack can teach computer science. Continue on for each subject.

Then let students go as fast as they want. I'd have finished some subjects of HS by the time I was in 9th grade, and others I might still be trying. ¯\_(ツ)_/¯

Where you'd still need to be "placed" is social development. Which I think should actually be a subject.

Well, I think one should be able to come up with a way of doing so on the fly just from knowing that the square root function is monotonically increasing, and knowing binary search? Of course, doing it another way might be more efficient.