'Students who use AI as a crutch don't learn anything'
english.elpais.com
english.elpais.com
The other day I decide to try ChatGPT 4o with canvas. For a solid year, I've planned to create some easy membership registration and booking system for this small club I'm part of - just simple stuff to book rooms in a building.
Well, to my absolute amazement - I had a working product up and running after 4 hours of working with ChatGPT. One block at a time, one function at a time. After a day I had built on a bunch of functionality.
So while I'm not completely clueless on back-end programming, my front-end skills are solidly beginner. But it felt like a breeze working with ChatGPT. I think I manually modified at tops 10 lines during all this, everything else was just copy/paste and upload source files to ChatGPT.
Any errors I'd get, I'd either copy/paste, or provide a screenshot.
I actually tried doing something similar when GPT3.5 came out almost two years ago, but it was just too cumbersome then. What I experienced the other day felt lightyears beyond that.
So, did I learn anything ? No - not really. But did it solve a problem for me? yes.
EDIT: But I will add, it did provide solid explanations to any questions I had. Dunno how well it would have worked if my 70 year old mom had tried the same thing, but a gamechanger for people like me.
You did learn something, by the way: you learned how to use modern tools. You didn’t do things most efficiently but it was more efficient than writing code without the help of ChatGPT.
Busy work is work that is assigned merely for the purpose of occupying one's time.
That's not the same thing as practice. We drill children in arithmetic not to keep them busy but because it turns out repeatedly solving multiplication problems is an effective way to teach children their times tables.
Exactly right. In terms of education, there generally seems to be a blurry line between was is considered learning and what is considered memorization. If you memorize your times tables, it doesn't mean you've learned multiplication for example... oftentimes the ability to memorize and recall things is opposed to learning, which means leveraging previous knowledge to solve something new.
In the case of AI, it usually presents facts and opinions simultaneously (something a calculator famously does not do, for example). Facts are memorized, opinions are learned. In all core studies it's always been more important to understand what you're solving for, and why, rather than "how" to solve it. The continued dissolution of the "how" barrier is a net benefit for all of civilization, and when experts of "why" are valued more than experts of "how" the world will be a much better place.
I was shit talking education if anything :)
The latter is even sillier. Your position might be indefensible.
Since AI can’t invent new stuff, who will do that? Juniors who haven’t learned anything because of those tools? Or seniors who will disappear one day because they are retiring or are being replaced by AIs?
I already work with juniors who use ChatGPT and cannot explain what they wrote. They have a fucking engineers degree and don’t know anything. It’s catastrophic and may increase in the future. What will happen if it continues like this?
My point wasn’t that writing code with AI is bad, my point was that writing code for the sake of writing code is bad. If something already exists, use it. If something doesn’t exist, build it, bring something new to the world — whether that’s with hand-typed code or ChatGPT assisted code, I don’t care.
I think we should write less code.
I don't disagree with this from a business perspective but for an engineers perspective I find it severely limiting.
Even very very basic things should probably stay fresh foe you. If you cannot implement a basic parser ( recursive dexent / pratt etc) you will very likely reach for regex when there is likely a better solution that isn't a lot of code.
You should probably know how to write leftpad... Or how to strip ascii whitespace using an ArrayBuffer and a for loop in JS. These are things that is extremely easy but a little tedius to do but are fundamental skill to building up more complex solutions later.
You should probably know how to build and reason about some more advanced datastructures in your language. Basic trees, directed graphs, trie. These are things that if they are second nature for you to implement you can come up with novel solutions to actually novel problems when they come up.
You also get an innate understanding of where the performance characteristics of certian algorithms and datastructures actually lay. Because big O doesn't always tell the full story...
It's far more important to know what you want to do rather than how to do it.
Sure, you (or an LLM) can probably find a package that can quickly search for a file in an extremely large filesystem.
I'm guessing that the authors of S3 didn't have that luxury when they were building out this service years ago, though. There are very few people on Earth that deal with exabytes of data, and prior art only gets you so far in this scenario.
The only way something like that can be built is by truly understanding CS fundamentals. Most people study CS to become a SWE. If programming gets reduced to maintaining prompts and optimizing here and there, then there is a real risk of this discipline eroding over time.
Sometimes you don't need a binary tree, you just need a O(n) linear search but someone who has never played with the actual low level datastructures has no idea when that matters, so in their mind a hashmap makes a ton of sense because searching is between O(1) and O(log n) depending on implementation. But in many cases a flat array will be significantly more performant and is a simpler implementation but in their mind a hashmap is the better solution.
Now it probably doesn't matter, but when it matters it's better to know the the answer.
That for me is the big distinction between software engineering and software development.
Plumbers don't need to be engineers, but there are times when you really need an engineer to design the plumbing system.
Strive to be the engineer, purely because you will enjoy the craft a lot more, and people recognise drive and ambition.
It doesn't matter if you are in the right place at the right time if you don't have the skills to back it up.
Granted if all you want to be is a plumber that pipes APIs together and lives a different life, by all means I encourage you to enjoy life. But don't make students believe thats all their is to the industry.
Also I've found telling it specifically where it's messed up is way more effective than just shouting at it to fix it after it's failed a second time. And sometimes you just need to manually fix it.
I wrote an entire library last weekend, then rewrote it on Monday when I realised I'd messed up. Two things I wouldn't have bothered to do without AI doing the coding.
I know how the important stuff works and I could pick my way through the JS, but glad I didn't have to write it. I mean, I just wouldn't have.
Let's just not lose the documentation on how to modify/improve the AI when needed...
Maybe that can be the job of a very select few. Fixing AI the way we fix robots for manufacturing.
Even if you ask it to think about and correct the underlying problems, it still generates buggy code, often with the same problems it was pretty decent at reasoning about.
i was testing a data transform function and needed to pull a couple of different types of data, so i had it give me a simple script to get stuff out of alpha_vantage, yahoo finance, quandl, and a few kaggle datasets, and those generally worked out of the box, or close to.
a few of those actually have code on their site for the API, but OpenAI made it faster. i would have gotten there eventually, but it saved me a couple of hours.
anything more complicated and it gets messy. its like a genie who grants you exactly what you ask for in the most literal sense, and only has a 90% idea what you're talking about to begin with.
You're actually better poised to learn it now if you care to, now that you have a component you care about that already works that you can work from. Of course maybe you won't, maybe having GPT there will indeed prevent you from ever learning it, I don't know.
And this is exactly the concern.
The tools are genuinely useful for some tasks. But unlike club organizers getting to DIY some hobby project for their club, students aren't yet being tasked to produce useful things in the best way possible. They're being tasked to do fairly rudimentary things so that they can learn some fundamentals by way of practice.
And likewise, in trades like ours, juniors are tasked to do useful things, but they're given affordance to deliver those things in ways that help them learn some fundamentals by way of practice.
Students and juniors who skip the practice are basically just trading their future expertise and readiness to accomplish trivial things that either don't or barely matter. Some of them may become the first generation of expert prompt engineers, accomplishing things in totally new ways in what amounts to a novel trade, but many of them are just going to be shooting themselves in the foot.
This is the exact same exercise as my first time slapping Dynamic Drive scripts together to customize EzBoard back in the 2000s. I didn't understand any of it at the time.
This style of learning is hands on. You learn a little bit about the shape of the problem before you sit down and learn the theory.
Not everyone learns by opening the book first. Some people like to get their hands wet. Introduction through practical osmosis can lead to a fertile appreciation for the theory.
Even if the learner is climbing a suboptimal hill, they're still learning the subject landscape and getting a sense of it. It's still a gradient.
The entire subject of chemistry is like this. They feed you lies and half truths for the first few years of your undergraduate career so that you develop a sense for things. The real model is far too complicated and scary to introduce.
Today I wanted to try to create a tool for a game: snapshot a picture and a program recognizes the clipboard event and does image recog things and gives me data. I had a working poc in 3 hours and learned nothing. (Tbf I knew what I wanted and how to do it in general terms so the process might be different for a beginner.)
They force us to admit that with 8B people in the world, many of the questions we have and tasks we pursue have already been approximated countless times. They reveal that much of what we do is not so original.
Understanding -- human or machine -- is something different, and enables invention/originality/reflection in a way that recent innovations are still not yet able to acheive on their own.
Importantly, though, students and juniors are specifically being assigned challenges that are already known not to be novel or inventive, which is why these tools can so easily do the work for them. But when when they let the tool do so, they sidestep the unique growth opportunity they were given in the first place.
(Keeping in mind that this is a much higher bar than what we consider understanding in humans.)
We're talking about getting some project or task done. It's a practical exercise by definition. Any learning experience to be had is going to be hands on, and for student/junior-level tasks, it's not going to be some product of knowing deep theory in the first place.
But the process of identifying the boilerplate that needs to be written, the process of manually entering it, the process of debugging your own code that you wrote, the process of scouring for examples and explanations, the process of being held accountable in a teacher or colleague's review, the process of discussing your experience of the task with someone who already understands it well... these all provide extra opportunities for hands-on learning that are short-circuited when having an AI put it together for you.
Yes, script kiddies and VBA/Excel junkies in the sales department have been slapping together programs they didn't understand for decades, and many people have now joined the industry thinking that they might secure a career as an "engineer" by following tutorials well and pasting StackOverflow snippets efficiently. And while some people who found themselves starting on that path have eventually come to transcend it and learn fundamentals more deeply, the "slap it together" mentality, the "find a tutorial" mentality, and now the "have a chatbot do it" mentality easily become quiet traps for people who don't realize that they need to actively transcend them at some point.
You can genuinely learn a lot about football by playing Madden on your couch, but if you don't get out on the field and actually play some games, your dreams of making it into the NFL are probably not going to pan out.
But if it is, then I think you’re trading it for a career of trial and error.
Regularly I watch people at work spend a week trying to solve a problem, but because I learned the fundamentals at some point in my past life, I am able to break down the problem, identify the root cause, and solve it quickly.
It's like enrolling for a Calc 2, cheating on all the homework to get an A, and saying "did i learn anything? No, but it solved all of these annoying homework problems for me!" Now when you have to take the 1st exam you're screwed because you didn't learn anything.
My hunch is that people who use the process you are describing will still get a massive leg-up in learning skills like web development.
Often it isn’t a choice between using AI-assistance to get some working vs spending 20 hours figuring it out from scratch: it’s a choice between getting somewhere with AI or not doing the project at all, because life is full of things to do that are more rewarding than those 20 hours of frustration.
Anecdotally, I’ve heard from a bunch of people who always wanted to learn software development skills but were put off by the steep initial learning curve before you see any concrete progress… and who are now building useful things and getting curious about learning more.
What makes the most difference in building to learn is the tiny steps you take to build. Printing hello world for the first time, changing it and seeing something else, using inputs for the first time to print hello, [variable], getting that image to animate across the screen. Each step becoming a great foundation for further curiosity, rather than turning your project into a black box.
In contrast, I've heard from a bunch of people who wanted to learn software development, but now don't see a point since AI can do it. Same with drawing. There's a large growing apathy towards learning skills I've noticed.
This is why most advocates for it don't do it from the perspective of learning. They do it from the perspective of building fast in the hands of those who already grasp the foundations.
If you use a template off GitHub, read that template, dig into bits of it you don't understand, make some changes to it and see what happens - then you'll learn plenty.
LLMs are amazing tools for learning if you're willing to put the effort in. If you just use their output without trying to learn anything from it you won't learn much from it at all.
When I have to navigate to somewhere I haven’t been before, I generally do not read a map, but follow instructions from some navigation software. As a consequence, I often don’t really know where places are, just the route I take to get to a destination. With GPS navigation, I do not get lost, but neither do I have much awareness of how locations are spatially arranged.
Such technologies seem to always be like this.
A potion which removes a difficult task, but also dulls the ability to do such tasks oneself.
It is like that one SMBC comic https://www.smbc-comics.com/comic/identity “ Humans offloaded memory to books, then thought to computers. Now, we're offloading our desires to the network. All that remains are basic bodily functions, which well offload in another generation or two. At that point, well just merge into one united entity so, it all works out.”
In my old age (60), I've gotten a little bit philosophical about this issue. I'm old enough to have pored through entire textbooks and manuals, e.g., BASIC, HyperCard, Turbo Pascal, MS-DOS (to name a few). But I can still ask myself at the end of the day:
So, did I learn anything?
Those things are all flawed, temporary creations of some individual, and are no longer useful. On the other hand, there are certain things that I've learned, and consider to be "fundamental," such as math, physics, and admittedly, music. Now a philosopher might correct me and point out that my choice of "fundamental" is arbitrary, but if nothing else, those things are long-lasting. The laws of physics that I'm capable of grasping haven't changed in my lifetime, nor has the technique of playing the double bass without injury.
Perhaps a thing you could do is sit down and decide what things you consider to be fundamental enough (relative to your interests) to learn on a deep level, and what things you can interact with on a superficial basis by letting AI take care of them for you.
> Q. You say that the best experts of the future will be those who make the most use of AI. Are people who are waiting to use AI making a mistake?
> A. I get it, it’s an unnerving technology. People are freaking out. They’re getting a sense of three sleepless nights and running away screaming. It feels like an essential threat to a lot of careers. I think if you’re a good journalist, the first time you think, “oh no.” But then you start to see how this could help you do things better than before.
There are a lot of white-collar jobs where LLMs do more harm than good because a 1/4 hallucination rate means you waste too much time on wild goose chases. I briefly thought GPT-4 was useful for finding papers given a description of the results - I “kicked the tires” with some AI research and was very impressed. But when I tried to find papers on animal cognition, about 75% of the results were fictional, though supposedly authored by real animal cognition experts. And GPT-4o is even worse! The tools are just not good enough for my use case; Google Scholar is far more reliable.
I just don’t understand the childish motivated reasoning behind assuming the skeptics are scared. Maybe if I spent “three sleepless nights” talking to ChatGPT I would be more enlightened.
That’s one of the many poorly documented traps of LLMs: trying to use them to find papers like that is a fast-track to worthless hallucinations. If that was one of your first experiments I can’t blame you for thinking this tech is “more harm than good”.
LLMs are terrible search engines… except for the times when they are great search engines!
Learning when and what to use them for continues to be a significantly under-appreciated challenge.
> LLMs are terrible search engines… except for the times when they are great search engines!
But note that what you said about finding papers was wrong, it works extremely well for AI research. The reason LLMs are useless to me across the board is that these unpredictable and arbitrary limitations apply to everything, not just search. "Learning when and what to use them for" is pure trial-and-error because it seems to amount to guessing what tasks the 3rd-party data contractors trained the LLM to solve.
I am not a Python or JavaScript developer, nor do I write code for extremely well-known libraries. I use F# for oddball projects (often analytics), and GPT-4 was utterly useless for F# codegen. My first experiments with GPT-3.5 showed that it would plagiarize hundreds of lines of public F# projects, including from my own GitHub, without any prompt engineering or trial-and-error - it was just blind plagiarism. GPT-4 isn't quite that bad, but it's still not even close to being good enough to help me - in particular it has no understanding of high-performance F#. I would be spending far more time auditing and optimizing its crappy code. And time spent writing code has never been the limiting factor in my F# development.
I also do some recreational mathematics on finite geometry and combinatorial group theory; GPT-4 was utterly useless here, even with CoT prompting, and even though it solved more complex graduate-level algebra problems without any difficulty. Of course, those problems were repeated and solved in dozens of graduate textbooks. My cute little groups, not so much. I believe CoT prompting is theoretically incapable of helping GPT here since the computational complexity is too high. What CoT prompting gives you is a bunch of insidious errors that take time and effort to unravel.
Otherwise there's nothing I do that would even conceivably benefit from an LLM: it can't play guitar, it can't play with my cats, and I would never use it to communicate with friends or family. I guess I could fill my brain with shallow subject knowledge about something, a few choice sentences. But I'd much rather understand something in depth by reading a book. I'm not too busy to read a book. Otherwise... maybe I could use LLMs to write polite no-thank-yous to unsolicited recruiters.
This tech truly has nothing to offer me. I think you are failing to understand that, as a Python developer who maintains one of the biggest Python web frameworks and writes a popular blog for general tech audiences, LLMs are unusually well-suited towards your use cases, due to reasons that will not extend to people working in more isolated corners of the world.
I'm very aware of that, and it's something I've been telling other people as well. If you are a software engineer who works primarily in languages that are well-represented in the training data (for me that's JavaScript and Python) you have a HUGE advantage available to you thanks to these tools, and you're possibly better positioned than any other profession to make the most of them.
I just dumped Whisper transcripts from three recent podcast appearances I gave into a Claude project and asked it "Find direct quotes where I emphasize how useful LLMs are if you write Python and JavaScript" and it found this one:
> "And that's great for me because the languages I use every day are Python and JavaScript and SQL. And those are the three languages that language models are best at."
Running these kinds of fuzzy searches against transcripts is one of the many non-programming uses I have for this stuff now. Here's that actual quote in the video:
This is less a question of which jobs benefit from AI in general and which don't than it is a question of tasks and specific tools.
ChatGPT is not a search engine, so if you're looking for existing documents it's a very bad choice. But I've found myself using Perplexity—an LLM-powered search engine—more and more often because it reliably turns up results that Google fails to turn up.
I suspect Perplexity is still also the wrong tool for scholarly articles, but that's not a fundamental limitation of the tech, it's just a question of the focus of the tools so far.
A lot of people are very invested, whether emotionally or financially or both, in this stuff not being a flop. There’s a lot of motivated reasoning going on. It is necessary to believe that the heretics simply haven’t seen the light yet - to question that gets too close to questioning whether there’s any light to see.
> A. Calculators also made us lazier. Why aren’t we doing math by hand anymore? You should be taking notes by hand now instead of recording me. We use technology to take shortcuts, but we have to be strategic in how we take those shortcuts.
An unpopular opinion I have is that most of the doomsaying about technology making us dumber is true. Yes, even back to Socrates. I won't say all, but I'd safely say a lot. What happened was that we developed tools, lost certain capacities without necessarily losing the capabilities that came with them, and redefined the level a normal human should function at. My only point is that people don't like to think that maybe they themselves are less intelligent—in many ways—than people who urinated outside and didn't know what the sky was. But I don't see how it could be any other way. When we say things like "I don't need to remember, I can write it down", and "I don't need to do arithmetic in my head, I'll let a calculator do it", or "I don't need to read the article, someone will explain it in the comments" we are accepting the consequences of that, good and bad.
For example, I doubt any website programmer knows the circuitry, assembly code, OS level calls, networking, etc, that make any webpage element do anything. Let alone can sit there and calculate any of the mathematical requirements needed to do any of that. But they know how to use an IDE and a framework like React.
All this is a long way of saying:
…on the shoulders of giants. AI is just the new tool needed for the next step up.
Perhaps the confusion comes from the fact that we often produce more complex things as we move up layers. It's then assumed that the people who made them must be more intelligent, but as I said, I don't think that's a fair assessment.
I would say the real measurement for intelligence here is how much of the abstraction layers you actually understand. In other words, can you move your cursor back down the stack and operate just as well as in the higher layers? Can you do this while unifying the complex interactions between each layer into a cohesive model? I've noticed that even AI tends to be pretty bad at this last step. It often takes prodding to get it to see the subtle errors often introduced when working with complex systems.
For example, card payments are a crutch. If you pay by card / phone everywhere and then out of the sudden you are to pay in cash, it becomes mildly challenging vs. if you are used to pay in cash you don't think about it. The brain is capable of a great deal of automation, performing learned actions is effortless. Unlike a calculator or a spreadsheet, the buyer is not doing anything, just buying. It's not a bicycle, it's a crutch. It simply atrophies the mental bandwidth. The mind becomes more lax, less sharp, when it does not engage.
Now imagine what it will do to people's brains when instead of thinking about solutions themselves, they will ask the AI for everything. Those neurons will atrophy and the person will be even less skilled to ask the AI the right questions than if they did not use the AI in the first place. I think the key will be a balance between doing the work yourself and delegating the stuff to AI, but it will be difficult to find that balance. Just like smartphones can be very useful but in the end are a net negative to society.
To a level much higher.
We stopped doing many repetitive, tedious things, but in return moved to things that are way more abstract and complex.
And that's happening everywhere. Even farmers are getting ever closer to being full on system architects.
Oh, you didn't learn to do quickly calculate square roots in your head? Instead you spent that time on learning about relativity in high school physics class.
By calling the people of the past smarter, you are really underselling the amount and depth of abstract though happening everywhere today.
The math examples always get me, because there's a level of understanding missing here that, yes, you should still have:
You may not be able to do the square-root yourself, but you should have a good-enough understanding of what a square-root is that if you enter into a calculator the square-root of a 6-digit number and get back another 6-digit number, you should know immediately that's wrong, and double-check you entered it correctly.
Part of the concern a lot of us have is people who not only do not have that understanding, but that think it's unnecessary to get that understanding, and skip over it entirely.
I can do math with a calculator, but if it is taken away?
I can feed myself with doordash, but if it is taken away?
I can program a complex web-scale app, but if all those tools are taken away?
What is left?
Somebody who will die fast.
Reliance on all of this is removing agency and resiliency. By the law of numbers, the planet still has people who know some of the fundamentals that make the existence of the rest viable.
But if it is taken away?
I no longer need to calculate in my head, walk 10 miles to get food, or wait for weeks to hear from my friends who moved across the country. Maybe it's making me dumber, lazier, weaker compared to my forerunners, but dumbest of all would be to ignore these advantages just because I'm no longer appealing to some guys idea of Ideal Intelligence.
The real issue arises when it becomes far too tempting to immediately turn to an LLM for an answer, rather than taking a few moments to quietly ponder the problem on your own, engaging and manipulating, exploring different angles, etc. This kind of abstract thinking is a craft that only improves with consistent practice and deliberate effort.
„The crutch is a dangerous approach because if we use a crutch, we stop thinking. Students who use AI as a crutch don’t learn anything. It prevents them from thinking. Instead, using AI as co-intelligence is important because it increases your capabilities and also keeps you in the loop.“
Another thing to consider is the motivation of companies like OpenAI. Their products are designed to be used as a crutch. Their money is in total reliance on the product.
This is especially demonstrated in essay writing.
Many students associate essays with busy work because the topics they're asked to write about are boring. When the typical assignment that's given is "read this boring ass book from the 40s that's been in the curriculum for decades without revisiting its application in today's world, then write a 1000-word essay on a topic that's been discussed to death that you couldn't give less of a shit about; points will be deducted for views that stray too far from the norm," then it's absolutely unsurprising that most students will shove this into ChatGPT and call it a day.
On the flip side, when English or composition teachers are forced to assign thess assignments knowing full well that it's a crock of shit, then it is equally unsurprising that they will feed GPT into GPT and call it a day.
Students that know how to learn and are actually interested in becoming better writers will find ways around this. Teachers who have the freedom to design their own curriculums will be more creative about the types of prompts they assign and the books they have their students read.
The common link between the two? Money, of course!
> Q. You don’t like to call AI a crutch.
> A. The crutch is a dangerous approach because if we use a crutch, we stop thinking. Students who use AI as a crutch don’t learn anything. It prevents them from thinking. Instead, using AI as co-intelligence is important because it increases your capabilities and also keeps you in the loop.
> Q. Isn’t it inevitable that AI will make us lazier?
> A. Calculators also made us lazier. Why aren’t we doing math by hand anymore? You should be taking notes by hand now instead of recording me. We use technology to take shortcuts, but we have to be strategic in how we take those shortcuts.
I could give a bunch of long (and probably butchered) stories of my sessions with her, but in short, she's expects to be spoon fed the answers without properly digesting why they are the answers. She turns to ChatGPT because it will just automatically give her the answers without forcing her to do any thinking, just copy + paste in order to tick off another assignment.
There's probably a middle ground where you could use ChatGPT properly, but I've had her reach towards ChatGPT once for what essentially amounts to spelling errors that she never notices herself or pay attention to errors telling her that she spelled something wrong. It's kind of hard to be supportive of a tool when the tool itself can essentially act like a replacement of your thinking cap and can't say no.
"Farmers today are much less skilled and knowledge than farmers 50 years ago!"
Currently, we can count on one hand with AI does, and have fingers left over.
It manipulates language (including formal language like computer code) with a modicum of logical reasoning, provides conversational access into the knowledge stored in a corpus of text, and generates/transforms images.
AI will currently not fix your leaky faucet, cook you dinner or take your dog for a walk.
My daughter is 10 and she is learning factoring, long division, and other things that a calculator does very well with. But she’s not allowed to use it at this stage because she can’t learn while using a crutch.
She’s also learning to write essays. She writes her essays then puts them into ChatGPT and asks for analysis, feedback, explanatory revisions. Then she revises the essay on her own without being able to refer back to the advice. This is using AI as a complement to learning and it’s been remarkably powerful. She can get feedback immediately, it’s high quality and impartial, and she can do it as many time as she finds useful. So, the fundamentals of learning don’t change no matter how powerful or different the tools become. But ignoring the tools because they can be used in place of learning if used in place of learning is dumb.
Which seems silly to me, but what do I know, I'm not a teacher. Nobody does long division in real life after K-12 school. It is not a useful skill to have, and it is not a useful concept to know. If I have to divide two numbers I just use a calculator like 99% of the humans on the planet.
Knowing what division is, and what it means to divide one number by another is valuable, but can you just teach that without teaching the mechanics of "divide the partial dividend by the divisor, then multiply the partial quotient by the divisor, and subtract from the partial dividend, extending to the next blah blah blah blah"? Are we really training the next generation for a world without electricity?
To my surprise, when I did pure maths at A level I found the same ideas applied to dividing one polynomial by another.
Of course as a mere memorisable algorithm a computer can also do this, so I'm not sure how useful it is even to pure mathematics, but there is (or was, 20 years ago) some use to the idea.
This was a problem even 20 years ago. I remember one instance from my physics class in highschool, one student's result had gravity going in the wrong direction. The teacher used it as an example for why should understand what we're doing and not to blindly trust our calculators.
Of course, leveraging technology to do the exact same thing you would do without the technology is a terrible lesson on many levels.
Is this legal? It might be, but I've never seen other sites do it so it seems dubious.
This is already happening with genz in college.
https://www.theatlantic.com/magazine/archive/2024/11/the-eli...
Or what about the attention span crisis?
Or the lack of technical skills in genz?
I seriously doubt this is a generational thing as you seem to be arguing.
Smartphones and social media didn't exist until 15-20 years ago and we're now seeing the consequences.
https://www.hopkinsmedicine.org/health/conditions-and-diseas...
A better formulation is perhaps “students who use AI to reduce work learn different things”. It’s easy for purists to say there’s no value whatsoever in learning to use a tool rather than learning to do the work.
But that’s a judgment about the value of what is learned, and it’s kind of dishonest slight of hand to substitute that opinion.
I don't think this is true. We learn a lot: Deference. Dependency. Entitlement. Impatience. Conformity. Distraction. Overconfidence. Intemperance...
If something "does the thinking for you" it has a much deeper effect than simply being a "crutch for the mind". It changes our relation to the world, to knowledge, motives, ambition, self-control...
"AI" is going to change our minds, but from what I've seen so far the outcome is a really quite awful kind of person, a net burden to society rather than a creative and productive asset.
There is no reason AI can't work like a tutor, the current crop is just a first take on the human-AI interaction problem. The motivation part can be solved by gamification and constraints - you need to earn a number of points by chatting the AI, and those points are reported to your teacher/manager. So a triad of student+AI+human coach would solve the motivation part.
We've let these tech companies distill learning and creating down to a mouse click.
That’s what they said when the calculator was invented. Out with the old, in with the new! Sorry but not sorry life was so hard before but we got AI to do the work for us now.
Each calculation is just one step, it’s up to the user to figure out which steps to take and how to chain them together. they might even learn that the whole thing would be faster if they could do some of those calculations their head.
like if you’re trying to figure out how much wood to buy for a deck, you’d still need to break the big problem down into those individual computations to do in the calculator. Unlike an llm where you could just ask it and it’d jump straight to a final answer
I don't think AI eliminated the application of concepts from learning. I think that has been eliminated enough due to the erosion of our public education systems. If we are not capable of critical thought with the information our own peers present us, why would it be any different when we seek it from AI?
I had this happen to me once while shopping, where I could immediately tell that three items costing less than £1 each should not come to a total of more than £3, but the cashier needed that explained to them.
(And that's aside from anything about asking LLMs to output in a format suitable for mechanical validation, which they can generally do).
On their own they would. We just don't see it very much because of external forces managing it - like teachers not allowing students to use them early in school.
I'm software engineering student. I had a phase year ago where I was using ChatGPT a lot, a lot more than I ever should have.
And it messed up with my brain a lot. I felt I became utterly lazy; to the point where quick fixes that should have taken me like, 10-15 seconds (?) I had to do with AI, which often took a very long time.
And the point of studying is to learn. You won't learn anything if you have someone else write your software for you.
Some of them had customer service jobs where they became utterly confused if the change was 99 cents and the customer gave them an additional penny.
It seems to have become the norm for young cashiers to be unable to understand. And if you try to explain, they'll insist "I can't change it now I've rung it through". Some seem to think the system keeps an exact record of the quantity of each individual coin (or they just don't even know where to begin to think about it).
It's like the difference between a language you are fluent in and a language you are tentative in. If you're fluent, you have to make an effort not to listen to somebody's loud conversation, or not to pay attention to a billboard. They intrude into your consciousness. There's never a situation when I don't do simple arithmetic when exposed to it. I don't have to consciously figure out what 4 times 9 is. Subjectively, the number just pops into my head when I see the question.
edit: If you can't do this with explanations of identities or related rates, etc., it's hard or impossible to follow any quantitative or especially probabilistic argument. Even the simplest ones. I think this results in people for whom arithmetic is difficult faking it by trying to memorize the words used during quantitative arguments without having any real understanding. Just sort of memorizing a lot of slogans and repeating them during any argument that shares similar words. I think discomfort with arithmetic ruins people politically (as citizens), so I really do think calculators are a problem.
I'd say yes to basic arithmetic; but I can't really use my own experience as a software developer who started off in video games to justify why a normal person needs to understand trigonometry and triangle formulas, any more than I can justify why they need to study Shakespeare and Alfred Tennyson over e.g. Terry Pratchett and Leonard Cohen — "I find it intellectually stimulating" is perhaps necessary, but certainly not sufficient, given there's more to learn than we can fit in a lifetime.
Conversely, understanding the deeper meanings behind the stuff people do read, perhaps Pratchett, is valuable — the use of speciesism between Dwarfs and Trolls used as a substitute for racism, etc.
It's not like rhetoric stopped with old classics; therefore it is better to learn how it works in newspapers or viral memes, as if you keep to Tenyson you're just going to pattern match it to 1800s writing style.
Modern writers including Pratchett make lots of references to this canon. Some people also learn that they enjoy Shakespeare in school, and they would never read a play like that otherwise.
That modern writers reference old works is true, but so too did Shakespeare, and we don't reference his cultural ancestors to understand his works.
> Some people also learn that they enjoy Shakespeare in school, and they would never read a play like that otherwise
That fine, but it's not the job of school. If it was, we'd also have mandatory watching of foreign language horror films.
What makes Pratchett more "relevant" than Shakespeare? I would posit to you that the same order of magnitude of people (or more) read and watch Shakespeare plays by choice as read Pratchett books. Let alone modern adaptations like West Side Story. I get that you prefer the one, but not everybody does.
English classes also have you read lots of other more recent and more modern works. It's not exclusively Shakespeare. I remember, for example, Death of a Salesman, 1984, and Slaughterhouse-Five being part of my high school's English curriculum.
Lord of the Rings is apparently in several English curricula now.
Comedies of all kinds (from Aristophanes to Pratchett) usually take a backseat to tragedies, bildungsromane, and other more "serious" genres in English classes because they tend to be much less timeless and much more current in terms of their themes and references.
> That modern writers reference old works is true, but so too did Shakespeare, and we don't reference his cultural ancestors to understand his works.
Schools do generally have you read ancient writers, often including Greek comedies and tragedies, and some have you read medieval British literature and those references do, in fact, come up. It's possible that you did not fully enjoy Shakespeare if you did not get that context.
> That fine, but it's not the job of school. If it was, we'd also have mandatory watching of foreign language horror films.
Believe it or not, if you take advanced French classes, you're going to watch a lot of French films and listen to a lot of French music for school! Since most students in English-speaking schools spend a lot more time studying English than French, they read a lot more English literature than French.
The language, the content, and the characters.
We don't live in a world where someone could fail to hear from their ship for a few weeks, thinking themselves unable to pay a weird debt, and have their loan shark's daughter dress up as lawyers and waltz into the trial to save the day. And we definitely don't encounter people who speak in Shakespearean English (even Victorian English is pushing it).
We do live in one where people try to commit insurance fraud as soon as they hear about it, a world where Magrats' have excessive candles (even if the magic isn't real), and a world of Vimes' Boots.
> Let alone modern adaptations like West Side Story
Or the Lion King. The adaptations are fine, the success make the connections to the world in which they are created.
> English classes also have you read lots of other more recent and more modern works. It's not exclusively Shakespeare.
I didn't say otherwise; I'm saying focus on those other things. I think the most modern thing we had was Ethan Frome.
> Believe it or not, if you take advanced French classes, you're going to watch a lot of French films and listen to a lot of French music for school! Since most students in English-speaking schools spend a lot more time studying English than French, they read a lot more English literature than French.
Again, that's fine for those learning French (and etc. for each other second language) to a higher level — given my experience with teaching myself a second language, and my self-tests over the years, there's a huge quality gap between what's a good grade in school and what's enough to get by with in practice — but that's not most people, and as there's exactly 18 years in the first 18 years of your life, my point is: what should be mandatory in schools? You've got limited time, what is the "you must" rather than "here's something you may like"?
We live in a 3d world, every human has to reason about geometry many times in their lives and trigonometry is the most useful and easiest of geometry math. Same reason we learn about volumes etc, everyone learns cos, sin, tan, its just 3 things it doesn't take a lot to learn that.
Outside hobbies and having worked as a game developer, I can't recall a single time I've needed them.
I suspect most learn them for a test them forget them forever.
Now, statistics, percentages, and compound interest, those are things people would do better to know, as those are the tools of advertising and money and are unavoidable in modern life.
They do teach those in middle school before they teach trigonometry, so that is already covered.
> And yet, outside my hobbies of "being a massive nerd", I've not needed the maths of trigonometry for at least a decade.
Lots of people craft things or work with schematics in some way, not just highly educated people, to them having the basics around angles is often very useful. People like mechanics, carpenters etc, that sort of math is useful to solve so many basic everyday problems in more handy jobs.
Not well enough, or people wouldn't get fooled so often.
> People like mechanics, carpenters etc, that sort of math is useful to solve so many basic everyday problems in more handy jobs.
And for 90% of the world, they're not.
Put those things into those schools.
The more you have memorized the more nimble your thinking is. If you have a large vocabulary you can effortlessly express yourself with precision while others are thumbing through a thesaurus (or these days asking an AI to “rewrite”).
If you know the history of something you can have more interesting perspective and conversations about it.
There is almost no situation where the person with a lot of memorized knowledge is at a disadvantage to the person who needs to look everything up or rely on tools to do the work.
Yes, it takes time, but learning is exponential, and overtime, the pace will increase greatly.
the AI generation is not going to know how to do anything other than type into chatgpt
at which point human progress ends and we start going backwards
Aggression against what? Yourself?
I think you show a tragic misunderstanding of technology and what it is doing in the world. It's not the work that it's doing for you. It's the living. Is it really your life you want a machine to take?
Nobody wants to "work". Henry David Thoreau said ,"There is no more fatal blunderer than he who consumes the greater part of his life getting his living." All good, no? But that's not what "AI", in the hands of exploiters (or even yourself, as a self-exploiter) is going to do to you. Technology is more "productive" but creates more, not less labour.
Better to heed Max Frisch who said, "Technology is the knack of so arranging the world that we don't have to experience it." Would you employ a machine to enjoy a music concert for you? To have sex for you or play games for you so you're not troubled by the effort?
Two of the three games I play on a daily basis largely play themselves, so... yes, actually. I still have plenty of fun watching them.
It's actually surprisingly fun, like you I was in the "why would I let a game play for me?" camp before. As they say, don't knock it until you try it.
Meanwhile, the third game I mentioned which is Fate/Grand Order explicitly does not have any autoplay functionality, and it too is very fun.
Not so much a game, but I get you.
Is "AI" a curiosity machine then? Our error is to think it can do "work".
What we're describing about games is fun, but for some kids in poor countries "work" is now generating bizarre images to post for ad engagement - and of course the most popular images are the most messed-up ones they can get past the censor. Like pulling the lever on a demonic slot machine.
Steroids. It’s not a perfect metaphor, But I think it’s useful. Two people are trying to gain muscle mass. They both have an ideal starting point. First person has a healthy diet, lots of exercise, and sleep. The second person has all of the same things the first person however they also taking growth hormones.
Lots of folks look at the two results and will see lots of different things. Beauty is in the eye of the beholder I suppose. If you think the end results of the work should yield sculpted bodies with larger than normal muscles… you might opt to use hormones. However, if you think sculpted bodies with larger than normal muscles looks unrealistic or just not your style/goal… you would probably opt for a more natural approach.
Both have their merits and could be described as “fit” despite their differences. folks may value one over the other. people might fantasize about looking like thor, but if everyone actually looked like thor, things would be weird. My two cents: Thor is fiction, and while we need fiction. Im not going to pretend that anyone should look like thor in order to be in shape or to be described as fit. If we allow ourselves to be fooled into thinking that it can be normal to look like thor, then we are doing something wrong. Fiction should not become reality.
Calculators arent giving you “kind of correct” answers.
Do you have a link or more info? Without further context, the 41% doesn't tell us the whole story; all we have is a numerator lacking a denominator. Did bugs per line of code go up, or down? Did # LOC produced after using AI go up/down? For all we know, the increase in productivity caused average bugs per line to go down, rather than up, which is contrary to the argument you're making.
Efficiency does not go up. In fact, the users claims that they are more productive do not stand up to scrutiny.
Code quality decreases.
This is why I don’t just believe what people say. It makes no sense to me that something I am observing is sneaking in little bugs on nearly every suggestion somehow saves time.
It logically cannot be true that reading output code to make sure it is doing what you want is significantly faster than just writing what’s in your head.
Then it is supposedly good for boilerplate, but like, code templates have existed forever. There’s been tools with prescriptive, deterministic output that handle boilerplate for you for ages. So again here, it makes literally zero sense that a non-deterministic output is actually handling boilerplate better than just setting your env up properly.
I’m willing to believe that those posting on HN about their experiences with AI coding tools are doing so in good faith and also are probably able to evaluate their own experiences faithfully. They aren’t gaslighting themselves, as most folks using them for work are assumed to be doing useful work with them over baseline of not using the AI coding tools, or they wouldn’t use them.
Good enough works. By that same token, those who are not having better than baseline experiences with AI coding tools are probably “holding it wrong”/using the tools in a manner that they are ill-suited for etc. Those devs are probably not able to assess their own performance as well as those who are more proficient at coding generally, even without using AI coding tools.
Basically, good programmers are better able to use programming tools, even AI coding tools that are somewhat buggy in their implementations and/or output. Using the wrong tool for the job or using the right tool improperly is not an indictment of the tool, but of the user.
A poor craftsperson blames their tools. A quality craftsperson accepts their tools for what they are and aren’t, and adapts their tools and adapts to their tools accordingly.
The tools in this case is consistently giving wrong results and measurably not living up to claimed efficiency. This is literally your business paying Microsoft to reduce efficiency and output a worse product than had you not done that.
Maybe AI will be there some day, but as of right now, using deterministic tooling is still unquestionably the king of productivity helping.
And it's beneficial to ban calculators for learning, which is the point of the article?