I'm sure computer science has improved in high school over the last (gulp) 20 years, but when I did variations of IT and programming lessons before university, it was bad. This was peak "you must Microsoft Office"-era. I've been involved in outreach for almost as long at this point. A lot of kids ask sensible questions like 'when do I ever need to use trig in real life?', because the examples in lessons and exams are so divorced from reality that it feels pointless.
I do think there is pedagogical value in showing where these concepts can be used practically and the advantage of LLMs is that you can transform the examples to what you're actually interested in. For example the Red Blob Games series on A* pathfinding are really good at showing how Dijkstra and graph traversal algorithms work, for a use-case (video games) that is appealing to a lot of nerdy people.
"When do I ever need to use trig in real life" is an interesting question, because it points out certain flaws in the way our society approaches education. One of those flaws is the one you pointed out: the examples we use are not very interesting.
But there's another flaw that gets overlooked most of the time, which is that we're raising kids to believe that "why are you teaching me something that you're not 100% sure I will need in my day-to-day life" is a sensible question, when it really isn't.
Outside of my 2-year stint in the game development industry, I never really needed most of what I learned about trigonometry in my day-to-day life. But that doesn't mean it wasn't useful.
Yes, we should make the subject matter more approachable to kids, but we should also try to shift the paradigm so that kids are more open to learning new things.
If the subject matter isn’t something the kid has a natural aptitude or interest in, and it’s not practical, and it’s not being taught in an unusually captivating way, why wouldn’t kids push back? I don’t blame them. I think adults should be able to justify why we’re using what boils down to the threat of force (if we’re honest) to make them sit in classrooms and listen to us.
I’m not going to use force against him. Threatening to take away the computer or tablet is generally plenty. I don’t think it’s unreasonable that a precondition to using these things is ‘go to school’.
Because they can not go to the department that uses trig in university. This is the basic design of society; people want to remind these kids they are not good enough forever.
All the text books I've ever seen had practical examples in them.
Like determining the height of a tree or a house simply based on trigonometry.
Your suggestion is interesting but I am not convinced that a student would be helped by aligning the examples with their interests.
I could see a student asking how trig relates to computer games and the example the LLM generates becoming much more involved.
I see no problem with the examples being boring. The people that developed these techniques had such fundamental problems to solve and the wonder to me is the human mind that came up with these
methods.
All this to say, maybe we lack appreciation for the fundamental sciences that underpin every aspect of our modern lives.
> All the text books I've ever seen had practical examples in them. Like determining the height of a tree or a house simply based on trigonometry.
The trouble is a lot of those practical examples fall into the, "why would I care category". I had a high school physics teacher who described his university antics, one of which included a funny story of a bunch of his friends climbing on top of each other to measure the height of a flag pole. I guess the profs got tired of dealing with students scaling flag poles because I was measuring the height of mountains on the moon at the same university a couple of years later. The thing is nobody really cares about the height of a flag pole, while only a few would care about the height of the mountains on the moon.
The reality is the interesting applications are much more involved. They either require a depth of thought of process or a depth of knowledge that isn't appropriate for a textbook question. Take that trigonometry in games example. The math to do it was in my middle school curriculum, but it becomes obvious that computer graphics is more than trigonometry the moment you try to frame it as an example. I had linear algebra in high school. That will take you pretty far with the mathematics, but it will also be clear that a knowledge of computer programming is involved. Even knowing how to program isn't going to take you all of the way because few are interested in rendering verticies and edges ...
And that is just the obvious progression of knowledge in a simple application. Physics itself involves buckets full of trigonometry in extremely non-obvious ways, non-geometric ways.
I think it's just clunky, like "a pencil can be used for a recipe". My first take is "wait, are we cooking a pencil? or stirring with it?"
The first meaning of "use for a recipe" is "use as an ingredient."
But then, it's a pretty weird thing to explain to begin with, approximately every human on the planet knows what the word "list" means. So what does this pseudo-definition add?
I have used the quiz-making learning tool within gemini. It is very good for things that would exist in a typical K-12 textbook. The first 30 or 40 multiple choice questions on a subject are usually pretty good and useful. But then it will tend to repeat multiple choice answers, give strictly wrong answers, repeat questions, or offer multiple valid answers. The answer explanations are what you'd expect with little human QA. Still a useful tool for people who sanity-check the given answers, but it might do more harm than good if people don't follow up on their confusion.
The Hawthorne effect[1] (aka "novelty effect") comes to mind. Even if students report finding this style of content more engaging after a single session, is that actually because it's better, or just because it's different?
If you ask it to explain some complex algorithm, it'll go "imagine a football field..." If you tell it you're a college-educated software engineer in the system prompt and ask it a non-cs question, it will go "imagine a variable..." instead.
It's a cute "how do I reach these kids?" idea -- find what they like and explain the concepts with custom-tailored analogues.
I don't think the failure mode here is really "7th graders will see through the superficiality of this really quick". I think the failure mode here will be:
> Explain computer science basics for a 7th grader interested in poop and butt-sniffing
Although who knows... maybe this will unleash a generation of memes of the likes we have never seen before. And if the side-effect is more people are at least conversant in more topics, well, maybe that's not a failure mode at all
But... which kids? Do we have a fundamental problem reaching kids who are interested in basketball? My kid had a period of being interested in dinosaurs, but I never felt the need to reframe everything in dinosaur-terms because of that. In fact, you kinda want them to broaden their horizons beyond dinosaurs?
The real challenges in education are elsewhere, and a lot of it has to do with socioeconomic status and bad influences early in life.
Okay, maybe I'm dumb, but I don't get how the first 3 are "simply plain wrong". They're open-ended enough that I have no trouble imagining how you could use those 3 data structures for those 3 purposes, so I must be missing some aspect of what you're trying to say.
It's not that they cannot be used. You can use any data structure for any purpose if you torture it enough. It's that the metaphors are terrible.
> "a list can be used for a recipe"
A recipe is not just a list of steps, it's also a list of ingredients, potentially an introduction, some pictures, etc.
Ask a kid to draw you a mock recipe, you won't just get a list of steps in return.
> "a set can be used to list all the unique ingredients you need to buy for a week's meals"
Ingredients have quantities attached. If I tell you to make a cake you need sugar, an egg and flour and give you all the steps but no quantities, you're not making a cake. A map is the obvious choice for storing ingredients.
I agree that ingredients are unique, but they have attached data which is just as relevant as the ingredient itself.
> "a map can be used for a cookbook"
I just don't understand how a cookbook is supposed to represent a map, it just doesn't make sense, not even with the additional context of the previous metaphors.
At best it would be somewhat understandable if it said a map can be used for a cookbook, with dish names mapping to recipes, but even this would be a stretch and assume a dish can be made in a single way.
Keep in mind the goal is to teach someone who has zero ideas about datastructures what they are, not to give some analogies to an experienced software engineer.
My beef with the list is what does "a list can be used for a list of steps" explain? Feels like the sort of thing you nod along too and then wake up the next day realizing you didn't learn anything at all.
Theres a reason why theres a grave yard of Google's dead projects.
Its annoying that software is such a high gross margin industry - I would love to see Googles cash get taken away so they cant take these vanity projects.
There's a _huge_ cottage industry of edu apps and programs that are both hugely expensive and not better at teaching, consuming a lot VC and end user $$$. As a first step, this seems not bad at all.
I do agree that it would be better to dial in on a pupil's interest than the grade level (my kid may be 7th grade in English but 9th grade in Chemistry, for instance.)
if the US government had done its job and split google's monopolistic ad business into what should be 4 different companies, then we wouldn't have this problem
Technology is a tool to expand the possible ways to educate, but isn't necessary for education to happen.
i.e. we've been educating people for 1,000s of years even without textbooks.
Education itself isn't primarily a technology problem. Treating it as such is an administrative failure, as is pursuing a technological solution in many scenarios that are first social in nature.
> i.e. we've been educating people for 1,000s of years even without textbooks.
And we've been doing a pretty crappy job educating people without written texts. The written word led to a tremendous acceleration of knowledge transmission. The printing press enabled that transmission at a larger, but unified, scale.
Anything we even remotely recognize as science has only ever been practiced by literate cultures.
Discarding technology for education because it's not a panacea is an absolute failure as educator.
> i.e. we've been educating people for 1,000s of years even without textbooks.
By using the tools available at the time we did, certainly. That involves physical tools like writing, but also non-physical tools like better ways of conveying and disseminating information, better ways of testing the efficacy of various approaches, etc, etc.
Education has to evolve, as it always has. While I'm not sure TFA is it, I do think LLMs will have a role to play in making learning more accessible and enjoyable for everyone, not just kids.
FWIW, I find the classic texts of certain fields much more intelligible than the intellectually shoddy 56th iteration of some overpriced glossy Pearson textbook. Compare a typical chemistry textbook with something like Pauling's "General Chemistry" which you can get from Dover, modulo any dated information. You will walk away with a far more solid grasp of the basic principles.
A lot of the failure of learning is a failure of teaching. Incompetent teachers throw disconnected information at you instead of trying to explain or lead you to an understanding of what something is about. I attribute part of this to a loss of solid philosophical coursework where you are taught to think from first principles, taught within a larger integral context, and taught to reason clearly. It used to be the case that everyone with a college degree had at least some basic philosophy under their belts (compare a Heisenberg to a Feynman to a Krauss; the progression is clear). And don't forget the success of the trivium and quadrivium or some variation of them that was often presupposed and prepared students for intellectual work.
I don't disagree with anything you're saying, really; education has been broken by things like too strong an emphasis on test scores, credentials, etc. This not only produces students who don't know how to learn properly, but those students then go on to become teachers who can't learn how to teach properly.
That said, teaching is hard. I don't fully blame teachers who cannot effectively convey subjects to 30 kids, especially these days. Even in an ideal situation, there's so much variance in how people learn best, that it would be hard to blame it on incompetence if a teacher cannot reach every one of their students.
Considering how hard it's going to be to fix the bigger problems with society* - obsession with credentials, lack of funding, better paying, less stressful jobs means less teachers, etc, etc - shouldn't we embrace tools that help kids learn things in a more accessible way to them? As I said, I don't think TFA is it, and we should obviously be aware of the issues, but surely people on HN of all places can see the value in tailoring subjects and lessons to a student's preferred method of learning?
* This is not to say we shouldn't also try to solve those problems
Technology can help, but in recent history, there's a track record of bogus pedagogy that insists on incorporating technology without any sound justification. Some of this was motivated by corporations trying to sell shit (like computers), some by silly or clueless teachers and administrators. Some of it is informed by dubious pedagogical methodologies like gamification.
For the most part, it's a matter of clear presentation, student engagement, and effort. A well-written textbook (many suck) and a good teacher (same) and a properly disposed student (which presupposes things like certain virtues; parents are responsible for teaching and supporting these for the most part). Technology won't get around the basic human reality, and sometimes, there's nothing to fix. Some people aren't interested.
Shameless plug but I made a similar tool called asXiv[1] which allows you to "ask" arXiv.org papers questions.
It also recommends questions on initial load that can help understand or explore the paper, here's a demo[2] from the popular Attention Is All You Need paper.
The code is all opensource[3], it uses the google 2.5 flash lite model to keep costs down (it's completely free atm), but that can be changed via env var if you run it locally.
Right. I do this all the time with Gemini. Add a pdf or a link and ask it whatever I want. It will even turn it into a podcast with two people discussing the entire paper that I can listen to on the tram to work.
whoa this is fantastic. wish I had known about this earlier! just made a similar product for reading arXiv / epub / pdfs called Ruminate (www.tryruminate.com), would love to hear what you think
I'm a former physics teacher, and while I'm impressed by the technology, I think this is a low efficacy innovation.
The real challenge in teaching Newton's laws of motion to teenagers is that they struggle to deal with the idea that friction isn't always there. When students enter the classroom, they arrive with an understanding of motion that they've intuited from watching things move all their lives, and that understanding is the theory of impetus: https://en.wikipedia.org/wiki/Theory_of_impetus
An AI system that can interrogate individual students' understanding of the ideas presented and pose questions that challenge the theory of impetus would be really useful, because 'unteaching' impetus theory to thirty students at once is extremely difficult. However, what Google has presented here, with slides and multiple guess quizzes, is just a variation on the 'chalk and talk' theme.
The final straw that made me leave teaching was the head of languages telling me that a good teacher can teach any subject. Discussions about 'the best pedagogy' never make any consideration of what is being taught; there's an implicit assumption that every idea and subject should be taught the same way. School systems have improved markedly since they were introduced in the nineteenth century, but I think we've got everything we can out of the subject-agnostic approach to improvement, and we need to start engaging with the detail of what's being taught to further improve.
> the head of languages telling me that a good teacher can teach any subject.
Tell me this wasn't foreign languages? :face_palm:
Okay, I was totally with you until this,
> but I think we've got everything we can out of the subject-agnostic approach to improvement, and we need to start engaging with the detail of what's being taught to further improve
I think if you walk into the bottom 80% of classrooms you would not see, interleaving, spaced repetition, recall-over-reread, or topic shuffling to avoid interference.
There's a load of understanding we've gained in pedagogy and human learning that has not affected how we structure formal education yet.
> I think if you walk into the bottom 80% of classrooms you would not see, interleaving, spaced repetition, recall-over-reread, or topic shuffling to avoid interference.
Where have you taught? I taught in Australia and the United Kingdom, where many of these things were mandated by the promulgation of spiral curricula by the relevant government departments. I'm aware in the US that, for example, algebra is taught as one or two block courses, but in the school systems I've taught in, algebra is taught as a few 'topics' of about a month in duration each, sprinkled throughout the whole four or five years in which mathematics is mandatory in secondary school. For Year 7 to 10 in Australia, there would be one or two topics for each of physics, chemistry, biology and earth sciences, covered across each year, building up from year to year. None of this was a choice by individual teachers or even schools; it was an artefact of the way the curricula are structured.
Right, that specific claim should be understood with US high schools and higher education in mind. There the massed approach rules the day.
Since these things I mentioned are well demonstrated to be effective and you don't think there's anything left to be had with a subject-agnostic approach, I infer you have a high opinion of how well these countries have implemented these "tactics". Is that right?
Yeah there seems to be some confusion. I agreed with the comment I replied to in the sense that teaching well requires specific domain knowledge and some specific pedagogy. Where I disagree is the assertion that the "tactics", to use your term, have been perfused through the system and there's nothing left to gain here.
He specifically says, "I think we've got everything we can out of the subject-agnostic approach to improvement"
So we all agree that subjects would benefit from specific interventions. The difference is he's going further and saying this is the only way forward; there are no general gains left to be had.
From the strength of the claim alone, this is hard to believe. Where do you stand on this?
Agree, it's unnecessarily limiting to say the well is entirely dried up re: improving status quo tools.
Charitable pov though, i'd say it's about leverage. Learning outcomes globally suffer steep steep cliffs and it's inevitably due to socioeconomic factors.
It's hard to argue that more chromebooks, spaced repetition, and catering to learning styles are the missing pieces johnny needs to get out of the hood.
as a person in tech i believed for a long time that if only we had better learning materials, people could orient and better self motivate around subjects. (learning needs to be hard. it's biology. brain takes notice and retains new and challenging stimuli. so "making learning easier" is a misnomer. the insight becomes how do we get people to self-motivate into hard things?)
I still think that's true, to your point, but all these takes are one of many many problems, and they aren't equal in leverage and i think that's where OP is coming from. there's outsized leverage in domain specific pedagogy.
The issue with learning things isn't that it hasn't been tailored to be interesting or relatable to me, it's just that it's a lot of content and it's hard. The solution is figuring out how to set up a type of spoon feed algorithm that checks that I'm understanding little bite size pieces along the way in addition to giving layman's terms for things that don't necessitate the formal description (e.g., deciphering math language).
ChatGPT Study mode has actually been quite good at this when you prompt it correctly and are studying a subject that it's well trained on.
Khan academy and brilliant are both excellent. They're hand crafted and limited in subjects and depth, but i think establish the current "roof" in how perfectly structured self-learning materials would look. I've heard from teachers using them in schools and found excellent results.
AI rephrasing words better to each individual isn't interesting to me. Automatic Interactive small quizzes, puzzles, and self adjusting difficulty level would be amazing, but i don't see AI really reaching that level.
When i see AI "quiz me on this" it gets stuck asking direct factual question about the text. But a good question challenges assumptions, and prod deeper understanding.
Here's a conversation using ChatGPT Study Mode I had a little bit ago covering Linear Algebra concepts that I wanted to learn. The concepts that were gone over in this chat aren't the most complicated, but I think you might find it interesting to look over since I think it actually does show we'll likely reach the level you're seeking. This conversation is with 4o shortly before the GPT-5.0 rollout, which is why it's a little less concise and more emotive.
Hush! Don't bring teaching experience into these discussions, you will trigger the EduTech people who have been promising a teaching revolution for 15 years already.
The 'teaching revolution' reaches much farther back than the last 15 years. It goes all the way back to the introduction of radio and television. It never did play out well.
I don’t have a firm opinion one way or another about that idea that a “great teacher can teach any subject” but it does bring to my mind the other idea that teachers are increasingly becoming “learning coaches” who aren’t only transferring knowledge into the students, but who are rather encouraging them to develop self-awareness about their own learning.
Why not a slippery surface? Simulations are less compelling (as they aren't the real thing and put the cart before the horse; the simulation presupposes physical laws, but doesn't demonstrate them) than a real example.
I'm not a teacher, but I think a simple change to "objects in motion stay in motion" could help with teaching it. Instead, tell students that any change in motion always has a cause, then ask them for the cause in different scenarios. Why does the ball stop rolling across the room? Why does the rocket launch into space? Why does the falling feather stop as it hits the ground? Then, ask what happens if there is no cause for change. Now you are left with the original law. That object will stay in motion.
My general experience with things like this from Google is to assume that this is at least one big step behind what they're doing now internally. Taking a position on how useful one finds this today effectively insulates from thinking more seriously about what could be done. If taken from a perspective of "what hints are laying around in this blog post or scientific articles about what's possible?" it's probably more effective use of time if you're going to invest time in reading it.
As an example, as you're reading it, try posing a few relevant counterfactuals.
>My general experience with things like this from Google is to assume that this is at least one big step behind what they're doing now internally
What they are doing internally after launching something like this is patting themselves on the back, updating their resumes, and promptly forgetting it exists.
No one who understands ai can rely on it to help us learn. I provided one with 100 citations I wanted to standardize and it deleted 10 and made up 10 to replace them. Can’t imagine this being used to replace a textbook or even explain a textbook.
But the problem is, you don't understand the passage, so therefore how will you vet the answers? Seems like hallucinations would be very very damaging in this use-case
I think your mileage will vary by subject and level.
If you’re a complete novice reading a niche graduate level textbook on Tolstoy’s critique of the Russian war effort in War and Peace, you’re going to get some wild hallucinations, and you’ll have no idea how to determine fact from fiction.
If you’re reading a high school textbook about the history of pre-revolution Russia, the models will have pretty comprehensive coverage of every concept you’re likely to come across.
Depends on the subject. For example in maths (assuming that one has a good background) you can verify the proofs yourself (and this isn't a given for highschool students).
I have also found another use for this. For example in studying modulation techniques in communication systems, I went back and forth between Monte-Carlo simulations and theoretical approximations to see how accurate each one is. And then added some more realistic error scenarios to do an end-to-end validation. In this case the LLM was used as a shortcut to write repetitive code that was verified manually, and this was complementary to the text-book, and made reading the topic more engaging, enjoyable, and comprehensive.
If you can't discern what good answers look like to the questions you're asking, you're not asking the right kind of questions.
Asking the right kind of questions is a genuine skill.
It applies to every domain of life where you are at the mercy of a "professional" or at the mercy of some knowledge differential. So you need to be a good judge of whether the answers you're getting are good answers or bad answers.
> Asking the right kind of questions is a genuine skill.
A skill we cannot rely kids to have, and which i think takes years of training and learning for even adults to really acquire. (to be clear, i'm not thinking about AI prompting. I 'm thinking about assumption breaking and understanding prodding questions the learner asks themselves and seeks answers for, to build and refine their mental models of something they learn)
That's absolutely not true. Kids get trained how to ask questions very quickly from a very young age. Good responses to those questions fundamentally shape the entire developmental journey for kids and extends to their academic abilities in school.
Because questions are fundamentally about knowledge differentials, which will always exist for individual human beings. We can't at any point know everything.
Know how to know what you don't know and get a good grasp of what it means to know in the first place.
A great question can compensate for a simple answer.
Kids can ask questions, but they rely on an experienced teacher to effectively answer.
Teaching someone effectively through answering questions, require the teacher through the students questions to build a model of the students model. To answer not only the question directly, but also the question that should have been asked instead.
A good end-of-chapter quiz doesn't check that a reader read the next. It asks questions whos answer rule out possible (or common) incorrect mental models the reader may have built.
A learner skilled in asking truly excellent questions, makes questions for which even a bad or simple answer rule out and refine their assumptions.
And that is a skill i doubt is ever truly mastered.
Its like the X Y. A great teacher answers X instead of Y. A great learner asks about X in the firstplace.
> If you can't discern what good answers look like to the questions you're asking, you're not asking the right kind of questions.
Whaaaaat? How does this work? If you're trying to learn a new topic, how are you supposed to recognize a good (and truthful) answer, whether it's from an LLM or instructor?
I would argue this isn't a fair comparison. There's a big difference between a fairly open-ended discussion about a topic both parties are at least somewhat familiar with, and someone trying to learn a new subject.
When learning from a source like a textbook, docs, or being instructed by a person, I do not expect the source of truth to lie to me, and verify everything they tell me.
Did we all just collectively forgot basic literacy?
You as the reader when you're reading anything are supposed to verify claims the author is making.
You never expect anything to be sources of truth.
That's why every textbooks either cites the sources or proves their work.
Very rarely do you have any textbook that's just a list of facts out of thin air. I don't think I've seen a single textbook, even bad ones, do this. They always cite their claims, or they show the logical steps to prove or justify a claim. Good textbooks make it easy to follow and clearly show their steps for the convenience of their readers.
Any good textbook seriously considers both the historic literature on their subject, presents the context of that literature, and establishes some kind of proof of work that synthesizes that in order support their claim.
This is always the case. This is how basic academic writing is done.
And it is the job of the reader to follow those citations, and to verify the claims. That's literally how our academic system literally works.
This. When you're reading a reputable textbook, you're not thinking all the time "wait is this true?". You trust the author to be correct and truthful.
Imagine being handed a textbook with a warning in the first page "10% of the facts here are made up (including this one). Good luck!"
I was in the middle of typing the same question. This is the part that worries me about Generative AI; far too many people seem to have forgotten that its prone to confabulation and telling the user what they want to hear.
Sure, but if the LLM tells you the jump from step 2 to 3 in a calculus problem is the use of l'hopital's rule, you should be able to figure out pretty quickly if it's a red herring or not.
I love learning new things, Khan academy got me all the way through college, and I use ChatGPT / Claude to help me study papers regularly. I got frustrated really fast. Here is an example:
It starts with just this sentence, followed by a quiz on that sentence:
> When we are born, we inherit our genetic makeup and biological features. However, our identity as human beings develops through interactions with others in society. Many experts in both psychology and sociology have described the process of self-development as a key step to understanding how that "self" learns to function within society.
Followed by the quiz:
> Question 1: Based on the provided text, what is a key difference in focus between psychology and sociology regarding self-development?
A) Sociology is concerned with inherited traits, whereas psychology is concerned with societal norms.
B) Psychology studies societal functions, while sociology studies individual identity.
C) Psychology focuses on genetic makeup, while sociology focuses on social interactions.
D) Both fields exclusively study the biological features inherited at birth.
I thought D makes the most sense, as nothing in the immediate text provides a more granular answers. But it's not D. It made me question my intelligence, maybe I misread the sentence, maybe I needed to read something else? Oh there is a button for the entire PDF, but then isn't the purpose of it to break down the PDF into chunks and ask me questions on what I'm reading?
I'm sure this is a fixable bug, but I was looking for the "provide feedback" button, there is none.
It isn't you. None of those answers are correct. Sociology studies societies and cultures; collective behaviors at different scales within different niches, etc. It's an LLM hallucinating again.
Ignoring that, NONE of those answers are correct. It wants "C) Psychology focuses on genetic makeup, while sociology focuses on social interactions.", but that's not a true statement. Psychology is absolutely not focused on genetic makeup.
I wonder how this will contribute to our current declining literacy rates in a social climate that's already rife with anti-intellectualism and isolation. Even if this worked well, it appears to be to be a step backwards.
Call me pessimistic, but this technology looks more poised to replace teachers in schools than supplement them.
Cultural obituaries are often premature, and the one for literacy is no exception. A nascent contrary impulse is emerging: readers deliberately turning to long-form works as a form of intellectual resistance. I’ve been working through Norman Lewis’s Word Power Made Easy and Tom Heehler’s The Well-Spoken Thesaurus, not just to expand vocabulary but to restore the sinew of productive speech.
That project led me to conscript AI as a private tutor. With custom instructions, ChatGPT and Gemini now surface new words and nudge my prose toward clarity, turning a vague fear of erosion into conviction. A dedicated subset of users will inevitably harness such tools to strengthen their expressive range and communicative precision.
Until recently, my writing rarely left emails and journals. Now, with AI as scaffold and sparring partner, I draft short stories from my own life and recast them in the voices of authors I admire. This feels less like a technology poised to supplant teachers, and more like the substrate for a renaissance in autodidactic education.
In that case the problem isn’t what technology we do or do not introduce. A society that values literacy isn’t going to be duped by a demo and a blog post. However a society which does not value understanding, expertise, or teachers will take every opportunity to shortcut them.
Going by the atrocious salary expectation for teaching (which i firmly place as one of, if not the most, fundamentally important jobs in modern society) i think we've already established where we land on this.
And note how i didn't mention a country; i think this is a widespread issue beyond country borders currently.
My only other frame of reference is China where being a teacher is a very respectable and competitive job with good wages. (Software engineering is way less prestigious interestingly.) They even have a national holiday for teachers. So it would be essentially unthinkable for a tech company to drop a demo that threatens teachers.
> teacher is a very respectable and competitive job with good wages
My frame of reference is europe and America. I suspect more of asia than just china mirrors that respect, as education feels nearly unhealthily emphasised.
The so called advanced world that prosper on innovation, design and R&D is skimpimg on the very thing that is supposed to be our advantage. Its frustrating.
I’m not sure this approach is the right one, but the problem resonates with me.
I vividly remember hitting some blocker (7th grade chem, 4th grade reading, 2nd grade dinosaurs), where I had a question that the teacher dismissed.
My mind was stuck (blocked) as it couldn’t get past the question I had, and in a public school setting, it wasn’t worth the time for the teacher to dive down the tangent (or they simply weren’t prepared).
My hope for LLMs in education, is that they can supplement traditional curriculums such that students can go “off the rails” while still being nudged back to the desired outcome.
- How do we know electrons “spin”?
- Why does that word behave differently than others (in English)?
- How big is a sauropod compared to a blue whale?
I’ve found that on my own journey through education, it’s these sparks of interest that drive towards deeper understanding, rather than surface level rote memorization.
TFA says: “What if students had the power to shape their own learning journey?”
In the context of nonfiction/textbooks, this is already possible!
I didn’t read “How to read a book”[0] until high school, but it opened the world for me on another silly blocker I had, which is that material should be consumed start to finish.
Hopefully with “AI” more students will learn that there are many paths towards understanding the world, and not just the curriculum in front of them.
Another anecdote: In university, I ended up taking Circuits 1&2 before Calc 4 (diff. eqs).
This was fantastic, because everything I learned about Laplace, Fourier, etc. had an immediate connection to another area of interest, which made the class much more engaging.
I was thinking about that when I bought my pinenote - imagine Tom Riddle's diary helping you with mathematics. Sadly I lost interest in the Linux side of the pinenote as it took a while to get to a working state and now I've got other things going on.
Does anyone remember the Person of Interest episode which saw the gang destroy a Samaritan-run startup to prevent it from usurping children’s education (via free tablets and AI powered teaching)?
I think that this is a really nice step on the way, but my vision is that it would generate the resource just-in-time as we're talking, based on my actual understanding and interests.
The "personalized" basketball example isn't a good one - your hand pushing the ball down and the floor pushing it back up aren't an action/reaction pair. You could, for example, push the ball down over a bottomless pit (or in space or whatever) so that you have one without the other. Nitpicky, yes, and it also seems plausible that this very example would be found in textbooks already, but I wouldn't like to use it.
Nor do I understand why you'd say "ouch, that hurt" while dribbling...
I'm excited by this multi-media approach. I've been avid self-learner for years, and I've often found the textbook format too dry, and forbidding, but ever since I started using NotebookLM, I'm diving into textbooks more and more. There is genuine value in creating a new format that meets learners where they are at.
I feel like this is thinking too small. I don’t want a better textbook. I want them to be basing this off of the experience of going to the most effective private tutors.
The textbook is there so that the model has something reasonably correct to work with and doesn't make up too much wrong stuff to teach.
For tutoring, I think the approach in https://www.nature.com/articles/s41598-025-97652-6 is promising. (Prompts are included in the supplementary material on the last page: https://static-content.springer.com/esm/art%3A10.1038%2Fs415... ) They start with an existing collection of worked exercises, give the chatbot access to the full solution and then let students interact with it to get a walkthrough or just check their own solution, depending on how much help the student thinks they need.
Plug for our https://uphop.ai/app : it's for adult learning / corporate training. We break down a desired job skill into small chunks, and engage the user with practice & give nuanced feedback. And of course like chatbots make it easy for user to ask more questions or go on tangents.
Would appreciate feedback!
There's a bit of overlap with Learn Your Way I guess. I'm not sure users need to toggle between alternate formats of the same instruction though. Instead the instruction itself should be as multi-modal as possible, and offer flexibility to ask questions... which even gemini.google.com offers so I'm not sure this is a net improvement over that.
This feels at first glance like another instance of an AI tool doing something humans already enjoy doing, rather than replacing work we'd like to avoid.
The teachers and professor I've known have always loved adapting their lessons to suit the interests of their students - I think that's a core educational instruction skill.
I'm open to hearing disagreements, but reading through the usages and evaluations does not leave me thinking of a tool that would provide any benefit greater than just giving teachers more resources would.
Also some stuff are just hard to learn. They require a good (even deep) familiarity with some foundational knowledge and it will be a slow process to go through those.
Isn't this just showing the effects of actively engaging the learner by placing a topic in contexts familiar/favored by learner versus just reading about a topic?
Like, if you had made the text pdf readers do some manual thinking by working on trying to place the topic into the same type of familiar/favored context, wouldn't that have been the better comparison?
I think using GenAI for learning is cool and exciting (especially for autodidacts) but I'm not excited by this particular study structure.
> I think using GenAI for learning is cool and exciting (especially for autodidacts)
I don't know. I've been trying, but I think there are two fundamental issues. First, I don't think it's all that useful for "out-of-order" learning and for explaining concepts in non-conventional ways.
To give you a practical example, there's a certain order in which we teach math, and every subsequent step builds on the previous one. But quite often, this order is just a matter of convention, not necessity. You can explain a ton of higher-math concepts in terms of high-school algebra and geometry, it's just not something we do because we don't intend to teach high-schoolers any of that, and for undergrads, it's more expedient to lean on their knowledge of calculus / mathematical analysis than to start by drawing triangles.
And not once have I succeeded in convincing an LLM to circumvent that. If a topic is explained using mathematical analysis in college textbooks, this is how it will always answer. Which actually sucks for that curious high-schooler.
But second, LLMs just aren't nearly as dependable as textbooks. It's not even the base error rate - I think they're 90%+ accurate on most run-of-the-mill scientific questions - but that once they make a confidently-sounding mistake and you try to drill down, they keep digging that hole and sending you more and more off track. It's amusing if you know the domain and can spot mistakes. It's a huge waste of time when trying to learn a new field.
It's precisely why vibe-coding is more useful to experienced developers who can immediately reject bad results.
"Out-of-order" and non-conventional explanations are interesting to consider. For both, I would expect LLMs to do poorly when there isn't much (or any) material for those approaches in the training data. My intuition would be the learner is going to have to be more exploratory via prompt engineering and still struggle against the tendency of the model to lean into classic or conventional explanations.
I don't particularly expect models to be dependable in responses, but I see how that presents a much larger problem in a learning context. I'm ok with bad responses that I can fight back against, but I also wouldn't reach for an LLM for a new field by default either.
For me, I do like using an LLM as a supplemental learning aid along with other traditional resources. I haven't tackled a deeper, new-to-me field yet with one. Maybe it's time for that . . .
Figure 10 of the technical report https://services.google.com/fh/files/misc/ai_augmented_textb... has a third n=30 group labeled "LCG" that is mentioned nowhere else. It scores 100% on "I felt like today's educational tool helped me gain a good understanding of the content." so I wonder what's up with that.
This is one of the areas where LLMs are really useful. At their core they aren’t “thinking” so much as transforming and reorienting data.
What’s been most valuable for me is the way they create a kind of imperfect but effective Socratic dialogue with whatever I’m reading. I was the kid who always had my hand up in class, not to show off, but because I hated leaving something unexplained. Good teachers could make a text come alive by answering those questions.
LLMs give me some of that back outside the classroom. Even when I ask them to speculate, the process forces me to interrogate the material and refine my own model of it. That’s changed how I read, learn, and even how I experience novels.
So innovation on this “Socratic interface” and other interfaces is pointed in the right direction.
As a side note, I'm going through all the Le Carré novels... it is a lovely experience to be able to ask an LLM more questions about 1960s British culture, West German Cold War politics, and Le Carré's background as a diplomat/spy. A lovely way to engage with novels.
Also, Smiley is getting up their in fictional characters I admire. Not Iroh level, but up there.
All of this hurts. GenAI cannot replace people grounded in reality. Especially not teachers and mentors. The effects will be nauseating to anyone who cherishes the development of human beings and minds in general.
Teaching and mentoring is a two-sided thing. The mentor, if adequately tutored or capable himself, learns more than the student. I understand that this is something "we" hope to achieve for AI but it's so insanely dumb to do it this early, it almost makes me angry. Almost. These people are just doing their jobs. So, as usual, I'm calling their bosses dumb fucking pathetic shitheads of idiot trash who fucking cost our kind sooooooo much fucking potential it almost makes me angry.
Sorry, gotta keep the anger at "almost", max. Can't be angry at lack of levels of consciousness or awareness. It's beyond genetic, a decision of "culture" and conformity; back to topic:
Teaching is one of those mythical edges that hone themselves. And I don't mean just the skill, I mean the entire category, the concept, the inter-generational action that happens on absolutely all levels. GenAI in education, applied anything below "correctly" on the scale of civilization, in any interval/integral, is taking XP points off a higher dimension that is exclusive to us in the kingdom of animals.
Don't fuck this up. Don't listen to your bosses. Quit. Found your own companies. I can't put it in words, yet. I don't have the peace of mind. But it's too damn important not to mention it. Every dimwit with access to drugs and hookers can make fuck-you money. For himself and others. FUCK THAT.
we could blame universities... but nobody forced their hand and they adapted to lower standards because ... [no peace of mind to take the time, ok, economists x'] ...
I can supply a personal narrative, though: ( I have to exclude the conspiratorial part )
I move back home when my little sister had issues in 8th grade. it was coincidental. I started tutoring her.
(background: After I turned 13, every spring, summer and fall, I WORKED at a Gaertnerei aka garden center until I finished my Abitur with a C)
She made progress. ( she's a bad example because someone fucked with her thyriod right when she started to make progress ( it's Germany, they are D.U.M.B with eugenics, they outsourced it to The USA because they are bad losers but whatever ))
She made progress. What did I learn? A LOT. Zero hand waving. In retro, I learned the MINIMUM ( I, do you understand who I felt like back then and NOW ? ) ( for clarity, I THINK I know where your comment comes from ) ... HOW can they fail to teach this to youngsters in University? Wait ... I get it. The system works well to reach it's objective, but CAN it be aware of what lies ahead? No way. I was 15, or 16, now 37, when I was planting some dumb trees in some dumb pot, when I realized, in my ADHD obliviousness, that neither my boss, nor his son, my age, my then colleague, knew how things would turn out. Working class, programmed* to be this way, ok.
Fast forward: my classmates, the Jahrgang before and after, all teachers for more or less ten years now. ... ... ... None founded their own school, despite abilities to get the finances, ... ... ... all abiding to the standard curriculum ... ... ... none of them being a Studienrat ( some higher regional rank ) ... ... all feeling very cool cuz they make 5k a month (if) ... ... ... ... ... .PERIOD, moving on:
THEY wouldn't think of
> supporting instructors or instructional evolution.
HOW THE FUCK WOULD/COULD COMPUTER SCIENTISTS, OR ENGINEERS THINK THE SAME?
GERMANY: (our region lacks E V E R Y T H I N G
... agricultural corporations with something worthwhile developing ... I had someone uniquely capable (and experienced in my Jahrgang) but ... ... ... nothing ... not even a 0.1%er now or soon to be ... so even if you judge form a punched drug dealer's aka Fortune 500 POV... N O T H I N G.... HOW, except portfolio capitalism featuring "please don't make my child understand how disrespectful my capabilities are to the evolution of my grandparents genes" ... )
I would be interested to see how capable it would be at going against cultural standards to preserve accuracy.
It's very hard for humans to imagine some mundane things when they go against their expectations. I would imagine AI would suffer similar issues, but there is a potential for them to do better. It strikes me as fascinating because I would consider it to be a sign of a greater level of reasoning, but because it goes against expectations would be considered worse by many.
Some examples of what I am thinking about.
Grass is only around 60 million years old, depictions of earth earlier than that frequently show grass.
Jesus did not grow a beard until a couple of hundred years after his death.
The point of the rigidity and uniformity of school is not that it is the best way to get everyone to understand everything if they try. The point is that it forces all the kids to try. School is not just for the most interested, it is for everyone.
No AI you ever create will get a kid to choose learning how math works over doing basically anything else with their time. The point of school is not to teach, it is to discipline children to participate in education. Otherwise, why have it at all? Kids can find extensive information and guides for basically any topic they want on the internet right now.
Whenever I see things promoting AI, the first question that comes to mind is;
"Who gets held accountable when the AI lies in harmful destructive ways?".
None of the people promoting these things address this question in any tangible or productive way, it is almost always yes its a problem, or dissembling.
Its not an uncommon or unworthy question. They just don't like the only possible answer preferring the the opaqueness is nobody gets held accountable because that is what the given incentives produce.
1. Much of this suggestion has nothing to do with AI. Offering learners alternative forms of media for a subject is interesting and probably useful, regardless.
2. About the AI part: Let's remember that at the other end of these bells and whistles there is a huge amount of expended electricity, and sprawling sever farm infrastructure. That's the hidden cost. For now, it might seem like someone else is footing the bill, but that will not last for very long - and in fact, it already starting not to last. See:
All the people at the forefront of AI really loved and thrived in highly academic settings from kindergarden to PhDs, their own lived experience doesn't match up with this product at all. Why are they making it?
EdTech has the worst returns of any industry in venture capital. Why?
There are no teachers who say that technology has generally improved experiences in classrooms, even if some specific technology-driven experiences like Khan Academy and Scratch are universally liked. Why?
When you look at Scratch, which I know a lot about, one thing they never do is allege that it improves test scores. They never, ever evaluate it quantitatively like that. And yet it is beloved.
Khan Academy: it is falling into the same trap as e.g. the Snoo. If you don't know what I'm talking about, it's about, who pays? Who is the customer? Khan Academy did a study that showed a thing. Kids are not choosing to watch educational YouTube videos because of a study. It is cozy learning.
But why does Khan Academy need studies for a test score thing? Why does Google? This is the problem with Ed Tech: the only model is to sell to districts, and when you sell to districts, you are doing Enterprise Sales. You can sometimes give them a thing that does something, but you are always giving them exactly what they ask for. Do you see the difference?
It doesn't matter if it's technology or if it is X or Y or Z: if the district asks for something that makes sense, great, and if it asks for something that doesn't make sense, or doesn't readily have the expertise to know what does and doesn't make sense, like with technology, tough cookie. Google will make something that doesn't make sense, if it feels that districts will adopt it.
We can go and try the merits of Learn Your Way, thankfully they provide a demo. All I'll say is, people have been saying, "more reading" is the answer, and there is a lot of fucking reading in this experience, but maybe the problem isn't that there isn't enough text to read. The problem is that kids do not want to read, so...
> EdTech has the worst returns of any industry in venture capital. Why?
I think this one is fairly simple. Half of consumer spending comes from the top 10% of earners, whose kids we can assume have generally pretty decent educations already. The people who need education help the most don’t have money to spend on it.
The parents who do have money to spend want to invest in tailored education from a human teacher, not cheap, generic scalable technology. So margins will be low.
So if you want to make money, you need to focus on things like enrichment and test/college prep for the top 10%. Helping inner city kids who are 3 grades behind in reading doesn’t print money and VCs don’t want anything to do with that.
> So if you want to make money, you need to focus on things like enrichment and test/college prep for the top 10%
But there's no potential market in the top 10%. I mean, these people just hire a good teacher and that's it. There's no room for improvement; there's nothing that can beat a good teacher.
> Helping inner city kids who are 3 grades behind in reading doesn’t print money
This is a political problem. Political problems cannot be solved by technological means. So there is no market here either.
Getting into an elite college is an arms race. Anything you can sell to a parent which will give them a leg up over other parents is a viable product. To put a finer point on it, a teacher + your product beats a teacher.
> Helping inner city kids who are 3 grades behind in reading doesn’t print money
100% and this is broadly why ed tech doesn’t move the needle.
> This is a political problem. Political problems cannot be solved by technological means. So there is no market here either.
Another POV is: pick your disruption.
AI stuff has definitely disrupted education... for the worse. It happened within a political and economic status quo. The AI stuff did not need to wait for the movement of any levers of power to happen at all.
If you are seeking a way to fix low returns in ed tech (and for that matter, Health IT, which is like, #2 worst performing sector): attack Enterprise Sales. Destroy it. Make stuff that destroys the monetization system where districts buy exactly what they ask for. It isn't complicated.
Scratch and early Khan Academy provide a template for good ed tech targeting the learner directly.
Whether you make $1 million or $1 billion doing this, I don't know.
Chegg got to, and fell from, great heights by delivering cheating, which ChatGPT does for free now. Cheating ALSO worked within a political and economic status quo, that 30% of students cheat, and that the cheating is a necessity, apparently, for the survival and thriving of a vast number of people, all around the world.
There are markets. Lots of them. You can do good or bad. Paul Graham doesn't invest in Cluely, even if it makes money, it's kind of evil (A16z doesn't care about cheating, the people who run it are the ones who cheated in school). So there are even opportunities that are missed by the very best seed fund.
To me, a big opportunity lies in things the government education cannot do. Some things good, some evil, some complicated. For example, no matter how hard it seems to try, the government cannot functionally collect on a trillion bucks of student loans. What does that mean for education? I don't know, but I think if you are looking for $1b+ opportunities, they're there.
If you have free weights, a bench, and a place to run, you are already 98% of the way to being a healthy fit human. There is ample information available on how to use those tools.
You don't need a trainer, a $10,000 gym machine, and a $5,000 stationary bike.
Education has gotten so insane with per-student spend, and the results are the same as the kids who had pencils and 10 year old text books.
You know what really motivates studying though? The promise of being completely useless when you finish studying since everything will be done by AI! What a motivator!
It's tough because the problems in education are so vast. Not that I'm saying you're wrong, but: everyone wants a stylistic answer to the question, "What is the problem with education?" Sometimes the style of that answer is malaise (you). Sometimes it is, some racist drivel. Which is pretty common on this forum, unfortunately.
Everywhere you look in education there are problems. There isn't going to be some stylized answer.
These Google guys - and a lot of other people who write comments online - go and promote something they think is a world view or theory, and is really just a bunch of stereotypes and projections of their own college-aged vengances. VC likes these kind of people! These Google guys fit that mold. I can agree with the broad strokes of techno-utopia, but that also means you need space to say that your app is bad, your art is ugly, and your text is long and boring.
These Google guys do not have space for criticism. They are Enterprise Sales. If the district asks for tasteless Corporate Memphis art, that's the art they're going to get - I'm going to focus on the art because I know something about art, and the text that appeared in the demo was so horrifically boring that I didn't read it. Have you opened a children's book? None of it looks like fucking Corporate Memphis!
One thing I am certain of is that these Google guys do have taste, they are smart people. Their problem is Enterprise Sales. Don't get me wrong. If you are narrowly focused on giving people what they want, your creative product will fail.
Seems very similar when Microsoft invested in Apple back in the day when Apple was about to die.
Their concern was that they would be the only one OS company standing and be defacto a monopoly and regulated such. So, it was a away to keep your 'weak' competition alive just enough not to make you the sole provider.
Steve Jobs was able to turn around Apple in such a fashion that they become even bigger by letting go of the PC market and going mobile.
I wish instead Google would instead find a good way to have exercise books, with:
1) well-thought out exercises (covering all cases, whether in math or Spanish)
2) CORRECT solutions (just saying because even ChatGPT gets it wrong even for high school math)
3) that you can enter them using pen (if need be on an iPad)
Just a way to make zillions of exercises if I want to. And for my kids, the problem is these days teachers won't (AND mostly can't, they just don't know their subject) help them make a lot of exercises.
Lot of exercises does not really help. What is valuable is feedback, not the chore itself. You either need someone more knowledgeable guiding you, or rely on insights.
I looked at the example for computer science basics for a 7th grader interested in food. Explanations include:
"a list can be used for a recipe"
"a set can be used to list all the unique ingredients you need to buy for a week's meals"
"a map can be used for a cookbook"
"a priority queue can be used to manage orders in a busy restaurant kitchen"
"a food-pairing graph can show which ingredients taste good together"
Maybe I'm over-estimating the taste of 7th graders, but I feel like I would get sick of this really quickly.
I'm sure computer science has improved in high school over the last (gulp) 20 years, but when I did variations of IT and programming lessons before university, it was bad. This was peak "you must Microsoft Office"-era. I've been involved in outreach for almost as long at this point. A lot of kids ask sensible questions like 'when do I ever need to use trig in real life?', because the examples in lessons and exams are so divorced from reality that it feels pointless.
I do think there is pedagogical value in showing where these concepts can be used practically and the advantage of LLMs is that you can transform the examples to what you're actually interested in. For example the Red Blob Games series on A* pathfinding are really good at showing how Dijkstra and graph traversal algorithms work, for a use-case (video games) that is appealing to a lot of nerdy people.
"When do I ever need to use trig in real life" is an interesting question, because it points out certain flaws in the way our society approaches education. One of those flaws is the one you pointed out: the examples we use are not very interesting.
But there's another flaw that gets overlooked most of the time, which is that we're raising kids to believe that "why are you teaching me something that you're not 100% sure I will need in my day-to-day life" is a sensible question, when it really isn't.
Outside of my 2-year stint in the game development industry, I never really needed most of what I learned about trigonometry in my day-to-day life. But that doesn't mean it wasn't useful.
Yes, we should make the subject matter more approachable to kids, but we should also try to shift the paradigm so that kids are more open to learning new things.
If the subject matter isn’t something the kid has a natural aptitude or interest in, and it’s not practical, and it’s not being taught in an unusually captivating way, why wouldn’t kids push back? I don’t blame them. I think adults should be able to justify why we’re using what boils down to the threat of force (if we’re honest) to make them sit in classrooms and listen to us.
I’m not going to use force against him. Threatening to take away the computer or tablet is generally plenty. I don’t think it’s unreasonable that a precondition to using these things is ‘go to school’.
> One of those flaws is the one you pointed out: the examples we use are not very interesting.
I think the truth is a lot simpler. Most kids won’t use trig in real life.
There was a popular joke during my studies in university:
Advanced math came in handy just once in my life. My keys fell in the toilet, and I realized the best tool was a wire bent like an integral.
Because they can not go to the department that uses trig in university. This is the basic design of society; people want to remind these kids they are not good enough forever.
All the text books I've ever seen had practical examples in them. Like determining the height of a tree or a house simply based on trigonometry.
Your suggestion is interesting but I am not convinced that a student would be helped by aligning the examples with their interests. I could see a student asking how trig relates to computer games and the example the LLM generates becoming much more involved.
I see no problem with the examples being boring. The people that developed these techniques had such fundamental problems to solve and the wonder to me is the human mind that came up with these methods.
All this to say, maybe we lack appreciation for the fundamental sciences that underpin every aspect of our modern lives.
> All the text books I've ever seen had practical examples in them. Like determining the height of a tree or a house simply based on trigonometry.
The trouble is a lot of those practical examples fall into the, "why would I care category". I had a high school physics teacher who described his university antics, one of which included a funny story of a bunch of his friends climbing on top of each other to measure the height of a flag pole. I guess the profs got tired of dealing with students scaling flag poles because I was measuring the height of mountains on the moon at the same university a couple of years later. The thing is nobody really cares about the height of a flag pole, while only a few would care about the height of the mountains on the moon.
The reality is the interesting applications are much more involved. They either require a depth of thought of process or a depth of knowledge that isn't appropriate for a textbook question. Take that trigonometry in games example. The math to do it was in my middle school curriculum, but it becomes obvious that computer graphics is more than trigonometry the moment you try to frame it as an example. I had linear algebra in high school. That will take you pretty far with the mathematics, but it will also be clear that a knowledge of computer programming is involved. Even knowing how to program isn't going to take you all of the way because few are interested in rendering verticies and edges ...
And that is just the obvious progression of knowledge in a simple application. Physics itself involves buckets full of trigonometry in extremely non-obvious ways, non-geometric ways.
> "a list can be used for a recipe"
I don't even know what it means, tbh. I feel it's going to confuse the hell out of 7th graders.
How is that difficult to understand? A recipe is an ordered list of steps of what to do. So of course a list can be used for a recipe.
I personally prefer a serious text without bringing in unrelated concepts like food, but this is still understandable.
I think it's just clunky, like "a pencil can be used for a recipe". My first take is "wait, are we cooking a pencil? or stirring with it?"
The first meaning of "use for a recipe" is "use as an ingredient."
But then, it's a pretty weird thing to explain to begin with, approximately every human on the planet knows what the word "list" means. So what does this pseudo-definition add?
I have used the quiz-making learning tool within gemini. It is very good for things that would exist in a typical K-12 textbook. The first 30 or 40 multiple choice questions on a subject are usually pretty good and useful. But then it will tend to repeat multiple choice answers, give strictly wrong answers, repeat questions, or offer multiple valid answers. The answer explanations are what you'd expect with little human QA. Still a useful tool for people who sanity-check the given answers, but it might do more harm than good if people don't follow up on their confusion.
The Hawthorne effect[1] (aka "novelty effect") comes to mind. Even if students report finding this style of content more engaging after a single session, is that actually because it's better, or just because it's different?
[1]: https://en.wikipedia.org/wiki/Hawthorne_effect
Gemini is extremely fond of this writing style.
If you ask it to explain some complex algorithm, it'll go "imagine a football field..." If you tell it you're a college-educated software engineer in the system prompt and ask it a non-cs question, it will go "imagine a variable..." instead.
It's a cute "how do I reach these kids?" idea -- find what they like and explain the concepts with custom-tailored analogues.
I don't think the failure mode here is really "7th graders will see through the superficiality of this really quick". I think the failure mode here will be:
> Explain computer science basics for a 7th grader interested in poop and butt-sniffing
Although who knows... maybe this will unleash a generation of memes of the likes we have never seen before. And if the side-effect is more people are at least conversant in more topics, well, maybe that's not a failure mode at all
> It's a cute "how do I reach these kids?" idea
But... which kids? Do we have a fundamental problem reaching kids who are interested in basketball? My kid had a period of being interested in dinosaurs, but I never felt the need to reframe everything in dinosaur-terms because of that. In fact, you kinda want them to broaden their horizons beyond dinosaurs?
The real challenges in education are elsewhere, and a lot of it has to do with socioeconomic status and bad influences early in life.
"a lot of it has to do with socioeconomic status and bad influences early in life."
Haha, you think most Googlers understand this? No chance.
This is why products like this fail, dead on arrival - the person leading the charge simply doesnt get it.
But hey go ahead and burn the cash of shareholders.
My issue with that it's more that those are extremely poor analogies.
The first 3 are simply plain wrong.
GenAI's gonna GenAI I guess.
Okay, maybe I'm dumb, but I don't get how the first 3 are "simply plain wrong". They're open-ended enough that I have no trouble imagining how you could use those 3 data structures for those 3 purposes, so I must be missing some aspect of what you're trying to say.
It's not that they cannot be used. You can use any data structure for any purpose if you torture it enough. It's that the metaphors are terrible.
> "a list can be used for a recipe"
A recipe is not just a list of steps, it's also a list of ingredients, potentially an introduction, some pictures, etc.
Ask a kid to draw you a mock recipe, you won't just get a list of steps in return.
> "a set can be used to list all the unique ingredients you need to buy for a week's meals"
Ingredients have quantities attached. If I tell you to make a cake you need sugar, an egg and flour and give you all the steps but no quantities, you're not making a cake. A map is the obvious choice for storing ingredients.
I agree that ingredients are unique, but they have attached data which is just as relevant as the ingredient itself.
> "a map can be used for a cookbook"
I just don't understand how a cookbook is supposed to represent a map, it just doesn't make sense, not even with the additional context of the previous metaphors.
At best it would be somewhat understandable if it said a map can be used for a cookbook, with dish names mapping to recipes, but even this would be a stretch and assume a dish can be made in a single way.
Keep in mind the goal is to teach someone who has zero ideas about datastructures what they are, not to give some analogies to an experienced software engineer.
My beef with the list is what does "a list can be used for a list of steps" explain? Feels like the sort of thing you nod along too and then wake up the next day realizing you didn't learn anything at all.
Maybe you need to ask it for 8th graders to get, "A Petri Net can be used for a recipe"?
This is a start, not the end.
Instruction and instructors won't be going away.
Most people have never looked at textbooks needing evolving.
It's like the LLM ai shift to not think about how software used to be.
yea this is stupid . agreed.
I don't know when these dorks will understand that education isn't a technical problem. Its a social and emotional problem.
existing material is clear enough to learn from.
Theres a reason why theres a grave yard of Google's dead projects.
Its annoying that software is such a high gross margin industry - I would love to see Googles cash get taken away so they cant take these vanity projects.
There's a _huge_ cottage industry of edu apps and programs that are both hugely expensive and not better at teaching, consuming a lot VC and end user $$$. As a first step, this seems not bad at all.
I do agree that it would be better to dial in on a pupil's interest than the grade level (my kid may be 7th grade in English but 9th grade in Chemistry, for instance.)
[Edit: fix typo]
if the US government had done its job and split google's monopolistic ad business into what should be 4 different companies, then we wouldn't have this problem
The average ROIC is doing a nice job in compensating for their awful marginal ROIC.
It’s both. Technology is a component (I’d we wouldn’t have books, recorded videos, multimedia aids, etc.).
Technology is a tool to expand the possible ways to educate, but isn't necessary for education to happen.
i.e. we've been educating people for 1,000s of years even without textbooks.
Education itself isn't primarily a technology problem. Treating it as such is an administrative failure, as is pursuing a technological solution in many scenarios that are first social in nature.
Nitpick: Language is technology, it’s not something we’re genetically born with and is critical for education to happen.
> i.e. we've been educating people for 1,000s of years even without textbooks.
And we've been doing a pretty crappy job educating people without written texts. The written word led to a tremendous acceleration of knowledge transmission. The printing press enabled that transmission at a larger, but unified, scale.
Anything we even remotely recognize as science has only ever been practiced by literate cultures.
Discarding technology for education because it's not a panacea is an absolute failure as educator.
> i.e. we've been educating people for 1,000s of years even without textbooks.
By using the tools available at the time we did, certainly. That involves physical tools like writing, but also non-physical tools like better ways of conveying and disseminating information, better ways of testing the efficacy of various approaches, etc, etc.
Education has to evolve, as it always has. While I'm not sure TFA is it, I do think LLMs will have a role to play in making learning more accessible and enjoyable for everyone, not just kids.
FWIW, I find the classic texts of certain fields much more intelligible than the intellectually shoddy 56th iteration of some overpriced glossy Pearson textbook. Compare a typical chemistry textbook with something like Pauling's "General Chemistry" which you can get from Dover, modulo any dated information. You will walk away with a far more solid grasp of the basic principles.
A lot of the failure of learning is a failure of teaching. Incompetent teachers throw disconnected information at you instead of trying to explain or lead you to an understanding of what something is about. I attribute part of this to a loss of solid philosophical coursework where you are taught to think from first principles, taught within a larger integral context, and taught to reason clearly. It used to be the case that everyone with a college degree had at least some basic philosophy under their belts (compare a Heisenberg to a Feynman to a Krauss; the progression is clear). And don't forget the success of the trivium and quadrivium or some variation of them that was often presupposed and prepared students for intellectual work.
I don't disagree with anything you're saying, really; education has been broken by things like too strong an emphasis on test scores, credentials, etc. This not only produces students who don't know how to learn properly, but those students then go on to become teachers who can't learn how to teach properly.
That said, teaching is hard. I don't fully blame teachers who cannot effectively convey subjects to 30 kids, especially these days. Even in an ideal situation, there's so much variance in how people learn best, that it would be hard to blame it on incompetence if a teacher cannot reach every one of their students.
Considering how hard it's going to be to fix the bigger problems with society* - obsession with credentials, lack of funding, better paying, less stressful jobs means less teachers, etc, etc - shouldn't we embrace tools that help kids learn things in a more accessible way to them? As I said, I don't think TFA is it, and we should obviously be aware of the issues, but surely people on HN of all places can see the value in tailoring subjects and lessons to a student's preferred method of learning?
* This is not to say we shouldn't also try to solve those problems
Technology can help, but in recent history, there's a track record of bogus pedagogy that insists on incorporating technology without any sound justification. Some of this was motivated by corporations trying to sell shit (like computers), some by silly or clueless teachers and administrators. Some of it is informed by dubious pedagogical methodologies like gamification.
For the most part, it's a matter of clear presentation, student engagement, and effort. A well-written textbook (many suck) and a good teacher (same) and a properly disposed student (which presupposes things like certain virtues; parents are responsible for teaching and supporting these for the most part). Technology won't get around the basic human reality, and sometimes, there's nothing to fix. Some people aren't interested.
Shameless plug but I made a similar tool called asXiv[1] which allows you to "ask" arXiv.org papers questions.
It also recommends questions on initial load that can help understand or explore the paper, here's a demo[2] from the popular Attention Is All You Need paper.
The code is all opensource[3], it uses the google 2.5 flash lite model to keep costs down (it's completely free atm), but that can be changed via env var if you run it locally.
1. https://asxiv.org
2. https://asxiv.org/pdf/1706.03762
3. https://github.com/montanaflynn/asxiv
That looks interesting! I put your Show HN (https://news.ycombinator.com/item?id=45212535) in the second-chance pool (https://news.ycombinator.com/pool, explained at https://news.ycombinator.com/item?id=26998308), so it will get a random placement on HN's front page.
This looks great. Just fyi this is also possible via alphaxiv's assistant. Navigate to a paper, click tools -> assistant.
[0] https://www.alphaxiv.org/abs/1706.03762
Seems legit, I'll have to try later. Just curious, why didn't you make this a commercial SaaS?
maybe because every LLM provider has an "attach file" feature so you can attach a paper and then ask questions about it?
what's the value add of the wrapper that this person wrote at all?
I built it because it was tiresome to save the pdf, open a new window, add it and give a prompt.
Simply replacing the domain arxiv.org with asxiv.org does all that for me now.
Also it links to pages in it's answers and scrolls the pdf to it on click, allowing to view the pdf side by side with the chat.
Right. I do this all the time with Gemini. Add a pdf or a link and ask it whatever I want. It will even turn it into a podcast with two people discussing the entire paper that I can listen to on the tram to work.
whoa this is fantastic. wish I had known about this earlier! just made a similar product for reading arXiv / epub / pdfs called Ruminate (www.tryruminate.com), would love to hear what you think
How is that different than generic RAG though? Genuinely curious
I'm a former physics teacher, and while I'm impressed by the technology, I think this is a low efficacy innovation.
The real challenge in teaching Newton's laws of motion to teenagers is that they struggle to deal with the idea that friction isn't always there. When students enter the classroom, they arrive with an understanding of motion that they've intuited from watching things move all their lives, and that understanding is the theory of impetus: https://en.wikipedia.org/wiki/Theory_of_impetus
An AI system that can interrogate individual students' understanding of the ideas presented and pose questions that challenge the theory of impetus would be really useful, because 'unteaching' impetus theory to thirty students at once is extremely difficult. However, what Google has presented here, with slides and multiple guess quizzes, is just a variation on the 'chalk and talk' theme.
The final straw that made me leave teaching was the head of languages telling me that a good teacher can teach any subject. Discussions about 'the best pedagogy' never make any consideration of what is being taught; there's an implicit assumption that every idea and subject should be taught the same way. School systems have improved markedly since they were introduced in the nineteenth century, but I think we've got everything we can out of the subject-agnostic approach to improvement, and we need to start engaging with the detail of what's being taught to further improve.
> the head of languages telling me that a good teacher can teach any subject.
Tell me this wasn't foreign languages? :face_palm:
Okay, I was totally with you until this,
> but I think we've got everything we can out of the subject-agnostic approach to improvement, and we need to start engaging with the detail of what's being taught to further improve
I think if you walk into the bottom 80% of classrooms you would not see, interleaving, spaced repetition, recall-over-reread, or topic shuffling to avoid interference.
There's a load of understanding we've gained in pedagogy and human learning that has not affected how we structure formal education yet.
> I think if you walk into the bottom 80% of classrooms you would not see, interleaving, spaced repetition, recall-over-reread, or topic shuffling to avoid interference.
Where have you taught? I taught in Australia and the United Kingdom, where many of these things were mandated by the promulgation of spiral curricula by the relevant government departments. I'm aware in the US that, for example, algebra is taught as one or two block courses, but in the school systems I've taught in, algebra is taught as a few 'topics' of about a month in duration each, sprinkled throughout the whole four or five years in which mathematics is mandatory in secondary school. For Year 7 to 10 in Australia, there would be one or two topics for each of physics, chemistry, biology and earth sciences, covered across each year, building up from year to year. None of this was a choice by individual teachers or even schools; it was an artefact of the way the curricula are structured.
Right, that specific claim should be understood with US high schools and higher education in mind. There the massed approach rules the day.
Since these things I mentioned are well demonstrated to be effective and you don't think there's anything left to be had with a subject-agnostic approach, I infer you have a high opinion of how well these countries have implemented these "tactics". Is that right?
I'd be curious to see how this is laid out in plan view. Are there any resources you can recommend on this topic?
You may not be wrong that tactics aren't sufficiently widespread, but that's the thing they're just tactics.
Spaced-repetition is a good example. It's so objectively better than other forms of memorization, but it's just one tactic for learning.
In this sense "teaching well requires a specific set of tools and tactics" is exactly how "a good teacher can teach anything" would make sense.
The problem is it doesn't make sense.
Yeah there seems to be some confusion. I agreed with the comment I replied to in the sense that teaching well requires specific domain knowledge and some specific pedagogy. Where I disagree is the assertion that the "tactics", to use your term, have been perfused through the system and there's nothing left to gain here.
He specifically says, "I think we've got everything we can out of the subject-agnostic approach to improvement"
So we all agree that subjects would benefit from specific interventions. The difference is he's going further and saying this is the only way forward; there are no general gains left to be had.
From the strength of the claim alone, this is hard to believe. Where do you stand on this?
Agree, it's unnecessarily limiting to say the well is entirely dried up re: improving status quo tools.
Charitable pov though, i'd say it's about leverage. Learning outcomes globally suffer steep steep cliffs and it's inevitably due to socioeconomic factors.
It's hard to argue that more chromebooks, spaced repetition, and catering to learning styles are the missing pieces johnny needs to get out of the hood.
as a person in tech i believed for a long time that if only we had better learning materials, people could orient and better self motivate around subjects. (learning needs to be hard. it's biology. brain takes notice and retains new and challenging stimuli. so "making learning easier" is a misnomer. the insight becomes how do we get people to self-motivate into hard things?)
I still think that's true, to your point, but all these takes are one of many many problems, and they aren't equal in leverage and i think that's where OP is coming from. there's outsized leverage in domain specific pedagogy.
Yeah, as a student, I have to agree.
The issue with learning things isn't that it hasn't been tailored to be interesting or relatable to me, it's just that it's a lot of content and it's hard. The solution is figuring out how to set up a type of spoon feed algorithm that checks that I'm understanding little bite size pieces along the way in addition to giving layman's terms for things that don't necessitate the formal description (e.g., deciphering math language).
ChatGPT Study mode has actually been quite good at this when you prompt it correctly and are studying a subject that it's well trained on.
Khan academy and brilliant are both excellent. They're hand crafted and limited in subjects and depth, but i think establish the current "roof" in how perfectly structured self-learning materials would look. I've heard from teachers using them in schools and found excellent results.
AI rephrasing words better to each individual isn't interesting to me. Automatic Interactive small quizzes, puzzles, and self adjusting difficulty level would be amazing, but i don't see AI really reaching that level.
When i see AI "quiz me on this" it gets stuck asking direct factual question about the text. But a good question challenges assumptions, and prod deeper understanding.
Here's a conversation using ChatGPT Study Mode I had a little bit ago covering Linear Algebra concepts that I wanted to learn. The concepts that were gone over in this chat aren't the most complicated, but I think you might find it interesting to look over since I think it actually does show we'll likely reach the level you're seeking. This conversation is with 4o shortly before the GPT-5.0 rollout, which is why it's a little less concise and more emotive.
https://chatgpt.com/share/68cc844a-14d4-8009-88e3-53f5d781b5...
Hush! Don't bring teaching experience into these discussions, you will trigger the EduTech people who have been promising a teaching revolution for 15 years already.
The 'teaching revolution' reaches much farther back than the last 15 years. It goes all the way back to the introduction of radio and television. It never did play out well.
You should have asked the head of languages to teach you tensor calculus.
I don’t have a firm opinion one way or another about that idea that a “great teacher can teach any subject” but it does bring to my mind the other idea that teachers are increasingly becoming “learning coaches” who aren’t only transferring knowledge into the students, but who are rather encouraging them to develop self-awareness about their own learning.
That concept of teaching seems like a nice idea in a private school setting but wildly incompatible with most schools in the Anglosphere.
Would simulation games work here?
Just giving the students access to something that simulates a frictionless world to play around with? maybe with a simple on off switch.
Something i've probably seen shared by others here in webgl at some point and far cheaper to run than genai
Why not a slippery surface? Simulations are less compelling (as they aren't the real thing and put the cart before the horse; the simulation presupposes physical laws, but doesn't demonstrate them) than a real example.
Kerbal Space Program Has entered the chat
I'm not a teacher, but I think a simple change to "objects in motion stay in motion" could help with teaching it. Instead, tell students that any change in motion always has a cause, then ask them for the cause in different scenarios. Why does the ball stop rolling across the room? Why does the rocket launch into space? Why does the falling feather stop as it hits the ground? Then, ask what happens if there is no cause for change. Now you are left with the original law. That object will stay in motion.
The issue is their intuition for the general case is actually gathered from a special case.
You need examples that point at the general case - like Newton's cradle.
Conservation of momentum helps.
My general experience with things like this from Google is to assume that this is at least one big step behind what they're doing now internally. Taking a position on how useful one finds this today effectively insulates from thinking more seriously about what could be done. If taken from a perspective of "what hints are laying around in this blog post or scientific articles about what's possible?" it's probably more effective use of time if you're going to invest time in reading it.
As an example, as you're reading it, try posing a few relevant counterfactuals.
>My general experience with things like this from Google is to assume that this is at least one big step behind what they're doing now internally
What they are doing internally after launching something like this is patting themselves on the back, updating their resumes, and promptly forgetting it exists.
*leaving, raising a round because they worked on this, promptly not doing anything without Google's distribution behind them
(see NotebookLM)
No one who understands ai can rely on it to help us learn. I provided one with 100 citations I wanted to standardize and it deleted 10 and made up 10 to replace them. Can’t imagine this being used to replace a textbook or even explain a textbook.
> explain a textbook
I've had very good luck using LLMs to do this. I paste the part of the book that I don't understand and ask questions about it.
But the problem is, you don't understand the passage, so therefore how will you vet the answers? Seems like hallucinations would be very very damaging in this use-case
I think your mileage will vary by subject and level.
If you’re a complete novice reading a niche graduate level textbook on Tolstoy’s critique of the Russian war effort in War and Peace, you’re going to get some wild hallucinations, and you’ll have no idea how to determine fact from fiction.
If you’re reading a high school textbook about the history of pre-revolution Russia, the models will have pretty comprehensive coverage of every concept you’re likely to come across.
Depends on the subject. For example in maths (assuming that one has a good background) you can verify the proofs yourself (and this isn't a given for highschool students).
I have also found another use for this. For example in studying modulation techniques in communication systems, I went back and forth between Monte-Carlo simulations and theoretical approximations to see how accurate each one is. And then added some more realistic error scenarios to do an end-to-end validation. In this case the LLM was used as a shortcut to write repetitive code that was verified manually, and this was complementary to the text-book, and made reading the topic more engaging, enjoyable, and comprehensive.
If you can't discern what good answers look like to the questions you're asking, you're not asking the right kind of questions.
Asking the right kind of questions is a genuine skill.
It applies to every domain of life where you are at the mercy of a "professional" or at the mercy of some knowledge differential. So you need to be a good judge of whether the answers you're getting are good answers or bad answers.
> Asking the right kind of questions is a genuine skill.
A skill we cannot rely kids to have, and which i think takes years of training and learning for even adults to really acquire. (to be clear, i'm not thinking about AI prompting. I 'm thinking about assumption breaking and understanding prodding questions the learner asks themselves and seeks answers for, to build and refine their mental models of something they learn)
That's absolutely not true. Kids get trained how to ask questions very quickly from a very young age. Good responses to those questions fundamentally shape the entire developmental journey for kids and extends to their academic abilities in school.
Because questions are fundamentally about knowledge differentials, which will always exist for individual human beings. We can't at any point know everything.
Know how to know what you don't know and get a good grasp of what it means to know in the first place.
Knowledge isn't absolute.
A great answer can compensate for a bad question.
A great question can compensate for a simple answer.
Kids can ask questions, but they rely on an experienced teacher to effectively answer.
Teaching someone effectively through answering questions, require the teacher through the students questions to build a model of the students model. To answer not only the question directly, but also the question that should have been asked instead.
A good end-of-chapter quiz doesn't check that a reader read the next. It asks questions whos answer rule out possible (or common) incorrect mental models the reader may have built.
A learner skilled in asking truly excellent questions, makes questions for which even a bad or simple answer rule out and refine their assumptions.
And that is a skill i doubt is ever truly mastered.
Its like the X Y. A great teacher answers X instead of Y. A great learner asks about X in the firstplace.
> If you can't discern what good answers look like to the questions you're asking, you're not asking the right kind of questions.
Whaaaaat? How does this work? If you're trying to learn a new topic, how are you supposed to recognize a good (and truthful) answer, whether it's from an LLM or instructor?
I would argue you're doing it right now.
By being skeptical of the answers, testing the answers, corroborating with other sources, etc.
This isn't new. This is literally how we've been exploring this knowledge game for thousands of years.
I would argue this isn't a fair comparison. There's a big difference between a fairly open-ended discussion about a topic both parties are at least somewhat familiar with, and someone trying to learn a new subject.
All knowledge is open ended...
I bet when you're learning a new subject you do the same exact thing.
When learning from a source like a textbook, docs, or being instructed by a person, I do not expect the source of truth to lie to me, and verify everything they tell me.
Did we all just collectively forgot basic literacy?
You as the reader when you're reading anything are supposed to verify claims the author is making.
You never expect anything to be sources of truth.
That's why every textbooks either cites the sources or proves their work.
Very rarely do you have any textbook that's just a list of facts out of thin air. I don't think I've seen a single textbook, even bad ones, do this. They always cite their claims, or they show the logical steps to prove or justify a claim. Good textbooks make it easy to follow and clearly show their steps for the convenience of their readers.
Any good textbook seriously considers both the historic literature on their subject, presents the context of that literature, and establishes some kind of proof of work that synthesizes that in order support their claim.
This is always the case. This is how basic academic writing is done.
And it is the job of the reader to follow those citations, and to verify the claims. That's literally how our academic system literally works.
This. When you're reading a reputable textbook, you're not thinking all the time "wait is this true?". You trust the author to be correct and truthful.
Imagine being handed a textbook with a warning in the first page "10% of the facts here are made up (including this one). Good luck!"
I was in the middle of typing the same question. This is the part that worries me about Generative AI; far too many people seem to have forgotten that its prone to confabulation and telling the user what they want to hear.
Sure, but if the LLM tells you the jump from step 2 to 3 in a calculus problem is the use of l'hopital's rule, you should be able to figure out pretty quickly if it's a red herring or not.
What you input along side the prompt can go a long way.
What is "it": what models did you try? What was your prompt? When did you try it?
I love learning new things, Khan academy got me all the way through college, and I use ChatGPT / Claude to help me study papers regularly. I got frustrated really fast. Here is an example:
https://learnyourway.withgoogle.com/scopes/1KNlGW5E/immersiv...
It starts with just this sentence, followed by a quiz on that sentence:
> When we are born, we inherit our genetic makeup and biological features. However, our identity as human beings develops through interactions with others in society. Many experts in both psychology and sociology have described the process of self-development as a key step to understanding how that "self" learns to function within society.
Followed by the quiz:
> Question 1: Based on the provided text, what is a key difference in focus between psychology and sociology regarding self-development? A) Sociology is concerned with inherited traits, whereas psychology is concerned with societal norms. B) Psychology studies societal functions, while sociology studies individual identity. C) Psychology focuses on genetic makeup, while sociology focuses on social interactions. D) Both fields exclusively study the biological features inherited at birth.
I thought D makes the most sense, as nothing in the immediate text provides a more granular answers. But it's not D. It made me question my intelligence, maybe I misread the sentence, maybe I needed to read something else? Oh there is a button for the entire PDF, but then isn't the purpose of it to break down the PDF into chunks and ask me questions on what I'm reading?
I'm sure this is a fixable bug, but I was looking for the "provide feedback" button, there is none.
This would be very frustrating to a student.
It isn't you. None of those answers are correct. Sociology studies societies and cultures; collective behaviors at different scales within different niches, etc. It's an LLM hallucinating again.
Ignoring that, NONE of those answers are correct. It wants "C) Psychology focuses on genetic makeup, while sociology focuses on social interactions.", but that's not a true statement. Psychology is absolutely not focused on genetic makeup.
> I'm sure this is a fixable bug
I'm sure it is, but the bar for accuracy in education is way too high for a mistake as blatant as this to be allowed to slip through.
(I'm sure someone will chime in about their useless Hum 10 teacher making an even bigger gaffe, and how this can be excused, it's just a beta...)
I wonder how this will contribute to our current declining literacy rates in a social climate that's already rife with anti-intellectualism and isolation. Even if this worked well, it appears to be to be a step backwards.
Call me pessimistic, but this technology looks more poised to replace teachers in schools than supplement them.
Cultural obituaries are often premature, and the one for literacy is no exception. A nascent contrary impulse is emerging: readers deliberately turning to long-form works as a form of intellectual resistance. I’ve been working through Norman Lewis’s Word Power Made Easy and Tom Heehler’s The Well-Spoken Thesaurus, not just to expand vocabulary but to restore the sinew of productive speech.
That project led me to conscript AI as a private tutor. With custom instructions, ChatGPT and Gemini now surface new words and nudge my prose toward clarity, turning a vague fear of erosion into conviction. A dedicated subset of users will inevitably harness such tools to strengthen their expressive range and communicative precision.
Until recently, my writing rarely left emails and journals. Now, with AI as scaffold and sparring partner, I draft short stories from my own life and recast them in the voices of authors I admire. This feels less like a technology poised to supplant teachers, and more like the substrate for a renaissance in autodidactic education.
In that case the problem isn’t what technology we do or do not introduce. A society that values literacy isn’t going to be duped by a demo and a blog post. However a society which does not value understanding, expertise, or teachers will take every opportunity to shortcut them.
Going by the atrocious salary expectation for teaching (which i firmly place as one of, if not the most, fundamentally important jobs in modern society) i think we've already established where we land on this.
And note how i didn't mention a country; i think this is a widespread issue beyond country borders currently.
My only other frame of reference is China where being a teacher is a very respectable and competitive job with good wages. (Software engineering is way less prestigious interestingly.) They even have a national holiday for teachers. So it would be essentially unthinkable for a tech company to drop a demo that threatens teachers.
> teacher is a very respectable and competitive job with good wages
My frame of reference is europe and America. I suspect more of asia than just china mirrors that respect, as education feels nearly unhealthily emphasised.
The so called advanced world that prosper on innovation, design and R&D is skimpimg on the very thing that is supposed to be our advantage. Its frustrating.
I’m not sure this approach is the right one, but the problem resonates with me.
I vividly remember hitting some blocker (7th grade chem, 4th grade reading, 2nd grade dinosaurs), where I had a question that the teacher dismissed.
My mind was stuck (blocked) as it couldn’t get past the question I had, and in a public school setting, it wasn’t worth the time for the teacher to dive down the tangent (or they simply weren’t prepared).
My hope for LLMs in education, is that they can supplement traditional curriculums such that students can go “off the rails” while still being nudged back to the desired outcome.
- How do we know electrons “spin”?
- Why does that word behave differently than others (in English)?
- How big is a sauropod compared to a blue whale?
I’ve found that on my own journey through education, it’s these sparks of interest that drive towards deeper understanding, rather than surface level rote memorization.
TFA says: “What if students had the power to shape their own learning journey?”
In the context of nonfiction/textbooks, this is already possible!
I didn’t read “How to read a book”[0] until high school, but it opened the world for me on another silly blocker I had, which is that material should be consumed start to finish.
Hopefully with “AI” more students will learn that there are many paths towards understanding the world, and not just the curriculum in front of them.
[0]https://en.m.wikipedia.org/wiki/How_to_Read_a_Book
Another anecdote: In university, I ended up taking Circuits 1&2 before Calc 4 (diff. eqs).
This was fantastic, because everything I learned about Laplace, Fourier, etc. had an immediate connection to another area of interest, which made the class much more engaging.
This is an underrated strategy
Looking forward to my copy of "A Young Lady's Illustrated Primer" a la Diamond Age
I was thinking about that when I bought my pinenote - imagine Tom Riddle's diary helping you with mathematics. Sadly I lost interest in the Linux side of the pinenote as it took a while to get to a working state and now I've got other things going on.
> imagine Tom Riddle's diary helping you with mathematics.
Your tone suggests that this wouldn’t be horrifying so I wonder what you meant by this.
Simply that it writes back and helps you along. hints might fade in after ten seconds of you staring at the problem.
I was referring to the visual experience, not the dark lord possessing people.
Does anyone remember the Person of Interest episode which saw the gang destroy a Samaritan-run startup to prevent it from usurping children’s education (via free tablets and AI powered teaching)?
Anyway …
Yes and Finch wasn’t sure if what team Machine had done was a good thing, which is interesting given he’s usually so sure about resisting Samaritan
I think that this is a really nice step on the way, but my vision is that it would generate the resource just-in-time as we're talking, based on my actual understanding and interests.
The "personalized" basketball example isn't a good one - your hand pushing the ball down and the floor pushing it back up aren't an action/reaction pair. You could, for example, push the ball down over a bottomless pit (or in space or whatever) so that you have one without the other. Nitpicky, yes, and it also seems plausible that this very example would be found in textbooks already, but I wouldn't like to use it.
Nor do I understand why you'd say "ouch, that hurt" while dribbling...
I'm excited by this multi-media approach. I've been avid self-learner for years, and I've often found the textbook format too dry, and forbidding, but ever since I started using NotebookLM, I'm diving into textbooks more and more. There is genuine value in creating a new format that meets learners where they are at.
I feel like this is thinking too small. I don’t want a better textbook. I want them to be basing this off of the experience of going to the most effective private tutors.
The textbook is there so that the model has something reasonably correct to work with and doesn't make up too much wrong stuff to teach.
For tutoring, I think the approach in https://www.nature.com/articles/s41598-025-97652-6 is promising. (Prompts are included in the supplementary material on the last page: https://static-content.springer.com/esm/art%3A10.1038%2Fs415... ) They start with an existing collection of worked exercises, give the chatbot access to the full solution and then let students interact with it to get a walkthrough or just check their own solution, depending on how much help the student thinks they need.
> I want them to be basing this off of the experience of going to the most effective private tutors.
Good goal, but they've got to start somewhere.
Delivering an education experience even 80% as effective as the best private tutors would be a huge achievement.
Plug for our https://uphop.ai/app : it's for adult learning / corporate training. We break down a desired job skill into small chunks, and engage the user with practice & give nuanced feedback. And of course like chatbots make it easy for user to ask more questions or go on tangents.
Would appreciate feedback!
There's a bit of overlap with Learn Your Way I guess. I'm not sure users need to toggle between alternate formats of the same instruction though. Instead the instruction itself should be as multi-modal as possible, and offer flexibility to ask questions... which even gemini.google.com offers so I'm not sure this is a net improvement over that.
How much of this is cutesy animations and how much is really valuable?
This feels at first glance like another instance of an AI tool doing something humans already enjoy doing, rather than replacing work we'd like to avoid.
The teachers and professor I've known have always loved adapting their lessons to suit the interests of their students - I think that's a core educational instruction skill.
I'm open to hearing disagreements, but reading through the usages and evaluations does not leave me thinking of a tool that would provide any benefit greater than just giving teachers more resources would.
Also some stuff are just hard to learn. They require a good (even deep) familiarity with some foundational knowledge and it will be a slow process to go through those.
https://learnyourway.withgoogle.com/
Link to an example:
https://learnyourway.withgoogle.com/scopes/AOcKkhsL/slides-n...
Upload your pdf so we can gather more training data but you’re providing it!
Isn't this just showing the effects of actively engaging the learner by placing a topic in contexts familiar/favored by learner versus just reading about a topic?
Like, if you had made the text pdf readers do some manual thinking by working on trying to place the topic into the same type of familiar/favored context, wouldn't that have been the better comparison?
I think using GenAI for learning is cool and exciting (especially for autodidacts) but I'm not excited by this particular study structure.
> I think using GenAI for learning is cool and exciting (especially for autodidacts)
I don't know. I've been trying, but I think there are two fundamental issues. First, I don't think it's all that useful for "out-of-order" learning and for explaining concepts in non-conventional ways.
To give you a practical example, there's a certain order in which we teach math, and every subsequent step builds on the previous one. But quite often, this order is just a matter of convention, not necessity. You can explain a ton of higher-math concepts in terms of high-school algebra and geometry, it's just not something we do because we don't intend to teach high-schoolers any of that, and for undergrads, it's more expedient to lean on their knowledge of calculus / mathematical analysis than to start by drawing triangles.
And not once have I succeeded in convincing an LLM to circumvent that. If a topic is explained using mathematical analysis in college textbooks, this is how it will always answer. Which actually sucks for that curious high-schooler.
But second, LLMs just aren't nearly as dependable as textbooks. It's not even the base error rate - I think they're 90%+ accurate on most run-of-the-mill scientific questions - but that once they make a confidently-sounding mistake and you try to drill down, they keep digging that hole and sending you more and more off track. It's amusing if you know the domain and can spot mistakes. It's a huge waste of time when trying to learn a new field.
It's precisely why vibe-coding is more useful to experienced developers who can immediately reject bad results.
"Out-of-order" and non-conventional explanations are interesting to consider. For both, I would expect LLMs to do poorly when there isn't much (or any) material for those approaches in the training data. My intuition would be the learner is going to have to be more exploratory via prompt engineering and still struggle against the tendency of the model to lean into classic or conventional explanations.
I don't particularly expect models to be dependable in responses, but I see how that presents a much larger problem in a learning context. I'm ok with bad responses that I can fight back against, but I also wouldn't reach for an LLM for a new field by default either.
For me, I do like using an LLM as a supplemental learning aid along with other traditional resources. I haven't tackled a deeper, new-to-me field yet with one. Maybe it's time for that . . .
Figure 10 of the technical report https://services.google.com/fh/files/misc/ai_augmented_textb... has a third n=30 group labeled "LCG" that is mentioned nowhere else. It scores 100% on "I felt like today's educational tool helped me gain a good understanding of the content." so I wonder what's up with that.
I would like AI that helps read and understand the text, but don't see real value having textbooks generated by AI.
Shocking they didn't have a printed book for a control, in addition to a digital reader.
Oh wow. This is actually a really bad idea and if your first thoughts are “oh wow this is awesome!” I don’t know what to tell you.
Shocking they didn't have a printed book control.
This is one of the areas where LLMs are really useful. At their core they aren’t “thinking” so much as transforming and reorienting data.
What’s been most valuable for me is the way they create a kind of imperfect but effective Socratic dialogue with whatever I’m reading. I was the kid who always had my hand up in class, not to show off, but because I hated leaving something unexplained. Good teachers could make a text come alive by answering those questions.
LLMs give me some of that back outside the classroom. Even when I ask them to speculate, the process forces me to interrogate the material and refine my own model of it. That’s changed how I read, learn, and even how I experience novels.
So innovation on this “Socratic interface” and other interfaces is pointed in the right direction.
As a side note, I'm going through all the Le Carré novels... it is a lovely experience to be able to ask an LLM more questions about 1960s British culture, West German Cold War politics, and Le Carré's background as a diplomat/spy. A lovely way to engage with novels.
Also, Smiley is getting up their in fictional characters I admire. Not Iroh level, but up there.
If Google didn't publish this it would come nowhere near the homepage of HN. Not exactly groundbreaking.
> The personalization pipeline
All of this hurts. GenAI cannot replace people grounded in reality. Especially not teachers and mentors. The effects will be nauseating to anyone who cherishes the development of human beings and minds in general.
Teaching and mentoring is a two-sided thing. The mentor, if adequately tutored or capable himself, learns more than the student. I understand that this is something "we" hope to achieve for AI but it's so insanely dumb to do it this early, it almost makes me angry. Almost. These people are just doing their jobs. So, as usual, I'm calling their bosses dumb fucking pathetic shitheads of idiot trash who fucking cost our kind sooooooo much fucking potential it almost makes me angry.
Sorry, gotta keep the anger at "almost", max. Can't be angry at lack of levels of consciousness or awareness. It's beyond genetic, a decision of "culture" and conformity; back to topic:
Teaching is one of those mythical edges that hone themselves. And I don't mean just the skill, I mean the entire category, the concept, the inter-generational action that happens on absolutely all levels. GenAI in education, applied anything below "correctly" on the scale of civilization, in any interval/integral, is taking XP points off a higher dimension that is exclusive to us in the kingdom of animals.
Don't fuck this up. Don't listen to your bosses. Quit. Found your own companies. I can't put it in words, yet. I don't have the peace of mind. But it's too damn important not to mention it. Every dimwit with access to drugs and hookers can make fuck-you money. For himself and others. FUCK THAT.
Personalized learning pre-LLMs has been a lot of hand waving and not really supporting instructors or instructional evolution.
it's brutal!!!!
we could blame universities... but nobody forced their hand and they adapted to lower standards because ... [no peace of mind to take the time, ok, economists x'] ...
I can supply a personal narrative, though: ( I have to exclude the conspiratorial part )
I move back home when my little sister had issues in 8th grade. it was coincidental. I started tutoring her.
(background: After I turned 13, every spring, summer and fall, I WORKED at a Gaertnerei aka garden center until I finished my Abitur with a C)
She made progress. ( she's a bad example because someone fucked with her thyriod right when she started to make progress ( it's Germany, they are D.U.M.B with eugenics, they outsourced it to The USA because they are bad losers but whatever ))
She made progress. What did I learn? A LOT. Zero hand waving. In retro, I learned the MINIMUM ( I, do you understand who I felt like back then and NOW ? ) ( for clarity, I THINK I know where your comment comes from ) ... HOW can they fail to teach this to youngsters in University? Wait ... I get it. The system works well to reach it's objective, but CAN it be aware of what lies ahead? No way. I was 15, or 16, now 37, when I was planting some dumb trees in some dumb pot, when I realized, in my ADHD obliviousness, that neither my boss, nor his son, my age, my then colleague, knew how things would turn out. Working class, programmed* to be this way, ok.
Fast forward: my classmates, the Jahrgang before and after, all teachers for more or less ten years now. ... ... ... None founded their own school, despite abilities to get the finances, ... ... ... all abiding to the standard curriculum ... ... ... none of them being a Studienrat ( some higher regional rank ) ... ... all feeling very cool cuz they make 5k a month (if) ... ... ... ... ... .PERIOD, moving on:
THEY wouldn't think of
> supporting instructors or instructional evolution.
HOW THE FUCK WOULD/COULD COMPUTER SCIENTISTS, OR ENGINEERS THINK THE SAME?
GERMANY: (our region lacks E V E R Y T H I N G ... agricultural corporations with something worthwhile developing ... I had someone uniquely capable (and experienced in my Jahrgang) but ... ... ... nothing ... not even a 0.1%er now or soon to be ... so even if you judge form a punched drug dealer's aka Fortune 500 POV... N O T H I N G.... HOW, except portfolio capitalism featuring "please don't make my child understand how disrespectful my capabilities are to the evolution of my grandparents genes" ... )
Google Veo 3 could also do a lot to spark interest by making 3D environments of Rome or Medieval Europe
I would be interested to see how capable it would be at going against cultural standards to preserve accuracy.
It's very hard for humans to imagine some mundane things when they go against their expectations. I would imagine AI would suffer similar issues, but there is a potential for them to do better. It strikes me as fascinating because I would consider it to be a sign of a greater level of reasoning, but because it goes against expectations would be considered worse by many.
Some examples of what I am thinking about.
Grass is only around 60 million years old, depictions of earth earlier than that frequently show grass.
Jesus did not grow a beard until a couple of hundred years after his death.
they definitely trained that voice on kahn academy guy
An exciting step closer to Neil Stephenson's Diamond Age.
The point of the rigidity and uniformity of school is not that it is the best way to get everyone to understand everything if they try. The point is that it forces all the kids to try. School is not just for the most interested, it is for everyone.
No AI you ever create will get a kid to choose learning how math works over doing basically anything else with their time. The point of school is not to teach, it is to discipline children to participate in education. Otherwise, why have it at all? Kids can find extensive information and guides for basically any topic they want on the internet right now.
The entire "AI education" thing misses this.
Well, if shoehorning interests in works for youth pastors...
Doesn’t work on mobile when you click into one of the examples. It says it is best for wide layouts
Whenever I see things promoting AI, the first question that comes to mind is; "Who gets held accountable when the AI lies in harmful destructive ways?".
None of the people promoting these things address this question in any tangible or productive way, it is almost always yes its a problem, or dissembling.
Its not an uncommon or unworthy question. They just don't like the only possible answer preferring the the opaqueness is nobody gets held accountable because that is what the given incentives produce.
Ask and answer. Its a worthy question.
1. Much of this suggestion has nothing to do with AI. Offering learners alternative forms of media for a subject is interesting and probably useful, regardless.
2. About the AI part: Let's remember that at the other end of these bells and whistles there is a huge amount of expended electricity, and sprawling sever farm infrastructure. That's the hidden cost. For now, it might seem like someone else is footing the bill, but that will not last for very long - and in fact, it already starting not to last. See:
https://www.newsweek.com/ai-data-centers-why-electric-bill-s...
A typical US household is paying a 26 USD "AI tax" this year through its electricity bill.
All the people at the forefront of AI really loved and thrived in highly academic settings from kindergarden to PhDs, their own lived experience doesn't match up with this product at all. Why are they making it?
EdTech has the worst returns of any industry in venture capital. Why?
There are no teachers who say that technology has generally improved experiences in classrooms, even if some specific technology-driven experiences like Khan Academy and Scratch are universally liked. Why?
When you look at Scratch, which I know a lot about, one thing they never do is allege that it improves test scores. They never, ever evaluate it quantitatively like that. And yet it is beloved.
Khan Academy: it is falling into the same trap as e.g. the Snoo. If you don't know what I'm talking about, it's about, who pays? Who is the customer? Khan Academy did a study that showed a thing. Kids are not choosing to watch educational YouTube videos because of a study. It is cozy learning.
But why does Khan Academy need studies for a test score thing? Why does Google? This is the problem with Ed Tech: the only model is to sell to districts, and when you sell to districts, you are doing Enterprise Sales. You can sometimes give them a thing that does something, but you are always giving them exactly what they ask for. Do you see the difference?
It doesn't matter if it's technology or if it is X or Y or Z: if the district asks for something that makes sense, great, and if it asks for something that doesn't make sense, or doesn't readily have the expertise to know what does and doesn't make sense, like with technology, tough cookie. Google will make something that doesn't make sense, if it feels that districts will adopt it.
We can go and try the merits of Learn Your Way, thankfully they provide a demo. All I'll say is, people have been saying, "more reading" is the answer, and there is a lot of fucking reading in this experience, but maybe the problem isn't that there isn't enough text to read. The problem is that kids do not want to read, so...
> EdTech has the worst returns of any industry in venture capital. Why?
I think this one is fairly simple. Half of consumer spending comes from the top 10% of earners, whose kids we can assume have generally pretty decent educations already. The people who need education help the most don’t have money to spend on it.
The parents who do have money to spend want to invest in tailored education from a human teacher, not cheap, generic scalable technology. So margins will be low.
So if you want to make money, you need to focus on things like enrichment and test/college prep for the top 10%. Helping inner city kids who are 3 grades behind in reading doesn’t print money and VCs don’t want anything to do with that.
> So if you want to make money, you need to focus on things like enrichment and test/college prep for the top 10%
But there's no potential market in the top 10%. I mean, these people just hire a good teacher and that's it. There's no room for improvement; there's nothing that can beat a good teacher.
> Helping inner city kids who are 3 grades behind in reading doesn’t print money
This is a political problem. Political problems cannot be solved by technological means. So there is no market here either.
Getting into an elite college is an arms race. Anything you can sell to a parent which will give them a leg up over other parents is a viable product. To put a finer point on it, a teacher + your product beats a teacher.
> Helping inner city kids who are 3 grades behind in reading doesn’t print money
100% and this is broadly why ed tech doesn’t move the needle.
> This is a political problem. Political problems cannot be solved by technological means. So there is no market here either.
Another POV is: pick your disruption.
AI stuff has definitely disrupted education... for the worse. It happened within a political and economic status quo. The AI stuff did not need to wait for the movement of any levers of power to happen at all.
If you are seeking a way to fix low returns in ed tech (and for that matter, Health IT, which is like, #2 worst performing sector): attack Enterprise Sales. Destroy it. Make stuff that destroys the monetization system where districts buy exactly what they ask for. It isn't complicated.
Scratch and early Khan Academy provide a template for good ed tech targeting the learner directly.
Whether you make $1 million or $1 billion doing this, I don't know.
Chegg got to, and fell from, great heights by delivering cheating, which ChatGPT does for free now. Cheating ALSO worked within a political and economic status quo, that 30% of students cheat, and that the cheating is a necessity, apparently, for the survival and thriving of a vast number of people, all around the world.
There are markets. Lots of them. You can do good or bad. Paul Graham doesn't invest in Cluely, even if it makes money, it's kind of evil (A16z doesn't care about cheating, the people who run it are the ones who cheated in school). So there are even opportunities that are missed by the very best seed fund.
To me, a big opportunity lies in things the government education cannot do. Some things good, some evil, some complicated. For example, no matter how hard it seems to try, the government cannot functionally collect on a trillion bucks of student loans. What does that mean for education? I don't know, but I think if you are looking for $1b+ opportunities, they're there.
It's like exercise equipment.
If you have free weights, a bench, and a place to run, you are already 98% of the way to being a healthy fit human. There is ample information available on how to use those tools.
You don't need a trainer, a $10,000 gym machine, and a $5,000 stationary bike.
Education has gotten so insane with per-student spend, and the results are the same as the kids who had pencils and 10 year old text books.
You know what really motivates studying though? The promise of being completely useless when you finish studying since everything will be done by AI! What a motivator!
It's tough because the problems in education are so vast. Not that I'm saying you're wrong, but: everyone wants a stylistic answer to the question, "What is the problem with education?" Sometimes the style of that answer is malaise (you). Sometimes it is, some racist drivel. Which is pretty common on this forum, unfortunately.
Everywhere you look in education there are problems. There isn't going to be some stylized answer.
These Google guys - and a lot of other people who write comments online - go and promote something they think is a world view or theory, and is really just a bunch of stereotypes and projections of their own college-aged vengances. VC likes these kind of people! These Google guys fit that mold. I can agree with the broad strokes of techno-utopia, but that also means you need space to say that your app is bad, your art is ugly, and your text is long and boring.
These Google guys do not have space for criticism. They are Enterprise Sales. If the district asks for tasteless Corporate Memphis art, that's the art they're going to get - I'm going to focus on the art because I know something about art, and the text that appeared in the demo was so horrifically boring that I didn't read it. Have you opened a children's book? None of it looks like fucking Corporate Memphis!
One thing I am certain of is that these Google guys do have taste, they are smart people. Their problem is Enterprise Sales. Don't get me wrong. If you are narrowly focused on giving people what they want, your creative product will fail.
Absolutely, the obvious best use-case for this technology -- from the ROI perspective -- is corporate HR training modules.
If anyone can take on textbook companies, large companies can.
Seems very similar when Microsoft invested in Apple back in the day when Apple was about to die. Their concern was that they would be the only one OS company standing and be defacto a monopoly and regulated such. So, it was a away to keep your 'weak' competition alive just enough not to make you the sole provider.
Steve Jobs was able to turn around Apple in such a fashion that they become even bigger by letting go of the PC market and going mobile.
Ah yet another project done to fill out promo doc
I wish instead Google would instead find a good way to have exercise books, with:
1) well-thought out exercises (covering all cases, whether in math or Spanish)
2) CORRECT solutions (just saying because even ChatGPT gets it wrong even for high school math)
3) that you can enter them using pen (if need be on an iPad)
Just a way to make zillions of exercises if I want to. And for my kids, the problem is these days teachers won't (AND mostly can't, they just don't know their subject) help them make a lot of exercises.
Lot of exercises does not really help. What is valuable is feedback, not the chore itself. You either need someone more knowledgeable guiding you, or rely on insights.