This will never take off. One, there is no money in Ed.Tech. There is no money in Ed.Tech. There is no money in Ed.Tech. What little money there is goes to the obvious stuff like student records databases. Anything that requires an ongoing subscription fee is dead in the water. The only reason those stupid smartboards took off is because they make school boards look cool, they are a one time cost, and can be paid for with bonds (because they are a one time cost). Teachers don't want them (projectors and document cameras are good, though). Ed.Tech is a wasteland of failed startups. Part of the problem is also that classic "the people with the purchasing power are not the people who will be using the product" problem.
Two, everyone outside of education thinks "well has anyone just tried sitting down with the kids and talking to them/explaining it to them?" Yes, obviously. The problem isn't that they are lazy, snot-nosed kids (that's a problem well within an experienced teacher's skill set to solve). The problem is what is the AI going to do with the kid says "fuck you" to the AI because they haven't eaten since lunch the previous day (school is the only place they get regular meals), or they don't even know what to ask because they are basically 4 grades behind in math, or the wifi is dead for the 8th time that month because the school board will never pay for infrastructure.
Three, what if the AI is just wrong and starts confusing the student? Even GPT-4 fabricates things all the time. Sure it can generally put words in grammatically correct order and is passible for writing no one is going to read anyway (like marketing emails). But the moment it requires actual domain knowledge all these AI models completely fall down because, again, they don't actually understand anything, they just are really good at guessing what word comes next.
Personalisation fails for a simple reason - people don't really want to learn.
Classes work because the teacher can inspire (or at least push), and there's peer pressure to learn (or at least keep up with the herd). I'm sure there's a dozen or so self-taught Python or Rust programmers here who will loudly refute what I'm saying, and point out that they were perfectly capable of learning something they were very interested in, but I bet a lot of them would also like to learn a foreign language or quantum physics and haven't gotten that done.
AI will fail for the same reason Youtube, DVDs, video cassettes, radio lessons, and phonographs all failed to be a revolution. If you want to learn something, and have the motivation, then reading the textbook is easily good enough (for theoretical subjects) and for less theoretical subjects you barely need a textbook, just lots of practice.
Yes, you can probably learn a little tiny bit faster with AI, if you (or someone staring over your shoulder) have the motivation to play some AI learning game rather than a more fun-optimised game that's purely about having fun, but it's a small optimisation.
Is this what we were missing in school? Our teachers dressing up with a tophat and bullshitting that they're lincoln and making stuff up?
But nothing like that exists today, as far as I know. I was disappointed with Khanmigo when I tried it back in August. Although I have a year's subscription to it, I've not asked my child to try it. It seemed too boring and told me my answer (to a problem it had given me) was wrong, even though I was right.
I'm optimistic, but I think there's work to do. I've not seen an AI system that can apply the pedagogy of George Polya: give the student a small hint by asking a question that could have occured to the student themselves.
"For tech CEOs, the dystopia is the point" - https://news.ycombinator.com/item?id=40371835
> a time-honored internet tradition: pointing out that the science fictional reference point a tech founder put forward was not an aspirational one, but, in fact, a dystopia containing a warning meant to be heeded, not emulated.
But students misunderstand in many different ways. How would it be tailored instead of another way of delivering lesson plans?
Even assuming you already had the topology of the problem space, you'd have to interrogate the student with all possible ways for each question to determine what's missing for them.
But part of the problem is impedance matching: failing students get little positive feedback because they have to correct many mistakes before they start getting things right. Interrogation makes the impedance-matching problem much, much worse (no matter how sweet and enticing the voice).
Sure, I believe in success patterns and strong leaders. But I'd need to see a sketch of how Khan understands and addresses the problems before I'd invest or rely on this as anything other than a stop-gap that continues the under-funding of mass education.
You don't need an LLM for this. Since the topic is specifically math, consider what that actually entails, especially for K-12 math education. The example is fractions. The rules for fraction problems are well-defined, we don't need an LLM to guess at the steps, we can literally write a simple program that encodes all the individual steps and prompts the student when they make a mistake to step back and walk through each small step (instead of taking big steps, like going from 1/2+1/4 to an answer have the student fill out each step like how to make the denominators the same and then adding the numerators and then, if applicable, simplifying).
Generating problems is also not something an LLM needs to be involved in. Such a strange and wasteful use of these tools.
I recently used ChatGPT to help me understand some Kubernetes and Terraform configuration. It was able to talk to me about the configuration and also help me understand some of the larger context- how information flows between the systems, what the terms mean, etc.
It provided direct explanations, and metaphors when I needed it, and let me ask clarifying questions.
This kind of learning system would be especially useful for people like myself who have Learning Disabilities, where once I understand a topic, I can speed ahead, but sometimes I'll just get "stuck", sometimes for days/weeks. Having a tutor can help, but there's a lot of embarrassment and that can lead to anxiety, which can in turn make it harder to learn.
Obviously such a system will need some safeguards around it, but having a system like this be able to both explain and point to primary sources could be a complete game changer for students.
at the same time, i see the limitations with this project in terms of what it wants to achieve.
tech alone is not the answer, AI or not. i had very positive experiences with teachers only using blackboards or whiteboards, and forgettable moments with those with the newest smartboards. to me, it is the teacher that brings out the value of these tools.
today's LLMs, no matter how large and expensive to run, do not capture the personal ingenuity that can replace teachers. this is not an old man shouting at the clouds opinion or any arrogance about humans being superior. critical thinking is a rare idea taught in schools, at least when i was growing up. these systems do not give an opportunity to do that.
i am happy to be proved wrong in the future but while i get the costs associated with researching and running the service, i feel it is unethical for them to be the ones charging for its access. it seems against the foundation's goals.
You can have all the knowledge at your fingertips!
What actually happened: ads shoved down our throats and way more distractions (porn, games, etc.)
Meanwhile math and reading levels continue to plummet.
So I don’t have high hopes. Those of us who use tools to succeed will continue to do so. Others who can’t control themselves will keep complaining. Nothing new.
AI will give us an answer with ads shoved into it with companies bidding to show their ad with the answer. People will do degenerate shit with AI, and gaming addiction will continue. People will still keep blaming the system and wondering why they can’t get ahead while spending most of their free time wasting it on gambling, porn, and games instead of learning.
The world will be split unevenly into people who research to contribute to AI knowledge base, and people who simply consume AI content.
The question is. How will we (the enlightened ones lol) make money off it?
The problem of education isn't the lack of resources it's the lack of purpose. The internet is all the resources you will ever need. Sure AI could augment this a little. Maybe it could re-explain something you can't quite wrap your head around. The problem is that kids know that the education system isn't even kind of preparing them for what's to come. Knowledge work and trades couldn't be further from what public school prepares you for. Having a teacher for every kid wouldn't change this. The problem is so fundamental, and the solution is so destructive that we can't just tack on one more thing to "fix" education.
Learning comes from doing, augmented with reflection. Public Schools don't ever do either of those things. The closest you ever get to doing anything that another human could want is building something in shop class. The closest thing you get to reflection is a grade that can only be negative feedback or the lack of feedback, plus there's no time for building on that reflection, just move onto the next thing.
He is classified as learning disabled, and has IEP through the school district which gives him accommodations including access to assistive technologies--such as an OCR reader that will read the words on a page to him to allow him to answer questions.
At the very least, AI for ed-tech is the next evolution of those technologies, which may even be allowed without the need of a 504 or IEP (e.g. each student in his class has a school issued laptop).
I'm also curious what subjects would be made obsolete because of the power of AI assistants. I certainly was not taught Latin, and my kids will never need to learn cursive writing.
If AI will be so revolutionary to me that first mean a complete overhaul of what students should learn? We haven't even properly done this reevaluation of the curriculum even for being constantly connected to the internet.
Students really need to learn a lot fewer things on average, but the fundamentals matter even more now. Honestly, 90% of the population doesn't need more than elementary math, but they really need to understand how percentages or pro-rata work.
In the end, looks like a puff piece by Bill Gates is still a puff piece.
I spent hundreds of hours poring over 1500 page textbooks and being given the driest lectures imaginable by people who were delivering them on autopilot.
I would estimate that one hour of khan academy was roughly equivalent to a weeks coursework, which was 3 lectures and probably an extra 2 hours of self study per lecture. It was undoubtedly the only reason I succeeded in the way I did.
[1] https://www.keplers.org/upcoming-events-internal/salman-khan
[2] https://www.penguinrandomhouse.com/books/740806/brave-new-wo...
Khan academy is an invaluable resource. When I have kids, it will be a core part of their curriculum!
About ten years ago I was thinking of returning to university and was required to pass a high school level math exam to prove I still could do math as an adult (I couldn't, really!). Khan Academy was a pretty incredible resource. I was in a time crunch so I also hired a tutor for two hours a week and the combination worked really well for me. I could see swapping out a real life tutor with an LLM being pretty amazing and much more affordable!
I didn't end up returning to school, but now I've gotten the AI bug and need to upgrade my math skills, and I'm thinking that using a LLM to help learn about foundational math that applies to LLMs would be pretty neat.
> Chapter by chapter, Sal takes readers through his predictions—some have already come true since the book was written—for AI’s many applications in education. His main argument: AI will radically improve both student outcomes and teacher experiences, and help usher in a future where everyone has access to a world-class education.
I'd argue that the very thing that made Khan Academy successful in the first place was not "accessible education" or "cost prohibitive" practices, but Sal's human attention to the students needs and all the amazing content he created.Thousands of people will pick up from this article that AI is the solution to the education, while in reality it is people like Sal who are the actual solution, with YouTube, AI and other tech just being the tools to help scaling human talent.
I am navigating through a textbook with the audio feature. I take photos of the pages with my phone, then use ChatGPT as my language partner to work through the problem sets. When i switch to the notebook, i complete the work and then take another photo of it and have ChatGPT verify my work.
I agree with Gates here -- if you are motivated this changes the education game.
I've sometimes wondered when I'll get called to the Dean's office because a student died in class because breathing wasn't specifically listed as a grading requirement on the syllabus.
AI education will continue this trend. AI can certainly teach you all kinds of remarkable skills. But the value of those skills will now plummet, because these are exactly the kinds of jobs that will be automated by AI!
the article goes on to describe how such an AI tutor might work:
"As you work through a challenging set of fraction problems, it won’t just give you the answer—it breaks each problem down into digestible steps. When you get stuck, it gives you easy-to-understand explanations and a gentle nudge in the right direction. When you finally get the answer, it generates targeted practice questions that help build your understanding and confidence."
reading it, and watching the demo of gpt-4o tutoring math (https://www.youtube.com/watch?v=dBrdd7xg-dg), I felt the opposite -- that if I had such a tutor, whatever motivation I had would have been stripped away by such a tutor. I really needed the quiet struggle with a problem, and feeling that I conquered it on my own, instead of being guided through a paint-by-number kit. I acknowledge that I might have an unusual learning style, or that I simply belonged to a generation that grew up without such tools, hence the aversion, much like my parents grew up without digital calculators (although they don't harbor any aversion to the use of calculators).
If the AI doesn't understand then you are just learning a script you are being fed.
Some automation is obviously useful for transmitting information about various topics but the way this is being framed would seem to undermine the very need to learn in the first place.
So I am curious, is Gates suggesting that we will overcome the fundamental issues here or is he saying that (which is kind of interesting in itself) it's plausible to have a technology that is fundamentally bad at something, be an effective teacher of that thing to others?
What Gates and Kahn describe is interesting, but it’s a little bit like talking about color theory and how the fact that a baseball bat is purple affected you after you just got whacked in the head with it.
(Khan Academy is excellent. Over a decade ago, my home-schooled daughter [we were not living in the US at the time] used Khan Academy throughout high school, scoring 2300 on the SAT and getting a full scholarship to a prestigious US college to study engineering.)
Education has a lot more edge cases than driving.
- kids give their assignments to AI so that it solves it for them, not for getting a hint.
- it has long been demonstrated human interaction is required to learn difficult things (such as reading/writing), otherwise kids would learn those skills on video.
- learning is about practicing difficult stuff. AI is about having the computer doing difficult stuff for you.
"Employees who can use AI effectively will be far more valuable than those who can’t."
The point of AI is that there is absolutely nothing to do for it to digest your problem. Show me someone who can't use AI? Must be illiterate?
"jobs of the future—which will become more enjoyable and fulfilling"
... finding and fixing the errors the AI made.
My parents are much better at counting with no calculator than my kids. Mental calculation is no more seen as a valuable skill now that everybody effectively has a calculator in his pocket (smartphone). AI will make some skills once valuable obsolete. The question is not how are we going to teach with AI but WHAT are we going to teach?
Google chatbot only serves google. Microsoft serves microsoft.
It will always cater to some agenda, it will have some biases. It will be nice, but it will play neatly their part in perpetuating corporatocracy.
As a unit you will have to play a long, or live in the woods without AI.
The fact is, Khan Academy was so useful to me in school precisely because it was NOT a resampling of a Wikipedia page, of every uninspired textbook I was ever forced to read. It was a smart, real life human being who understood the nuances of how actual human beings understood concepts. Sal didn't just break down complex mathematical subjects for me such that I was able to solve linear equations in echelon form, there was also a particular way he emphasised the words he spoke so I understood why something was important. The human factor is so important, and we're fucking over our next generation for market hype.
A few years ago, I was bullish on online learning. But no-one with common sense is as the helm. I am incredibly concerned that Microsoft is believing its hype at the cost of something that was at one point incredibly valuable to society.
I remember when MOOCs were the future of education, as students across the world would have free access to the best lectures from the best teachers and could watch them over and over again at their own pace. That ... didn't result in the second coming of Christ, to put it politely.
This part confused me on several levels.
Let's break it down.
---
1. Tutors (human) can not replace teachers on a mass scale
True, and the article touches on this: human tutors simply don't scale!
2. Tutors (human) should not replace teachers
False. Bloom's 2-sigma problem [0] states that on average, 1-1 tutored students outperform 98% of classroom-educated students. (!!!)
Therefore, if we could replace 1:30 education with 1:1 education, the evidence very strongly says that we should.
In other words, the only reason we have teachers, instead of tutors, is because we don't have enough tutors. If we had enough tutors, we would simply 1-1 tutor everyone -- assuming our goal is educational achievement.
3. AI tutors can never replace human tutors
This is a strange claim. It seems to imply that all technological progress will suddenly cease after today -- not just in AI, but human-computer interfaces, robotics, etc.? (Note: the article doesn't make this specific claim, so perhaps I'm straw-manning here.)
I don't expect convincingly human androids for a while. But are androids really necessary for teaching math? GPT-4o seems to do fine! [1]
---
As far as the tech goes, I think the only thing missing is memory and personalization. Currently GPT doesn't know what I know, and what I don't know. What is the dependency graph of knowledge? (KhanAcademy had one, but they've taken it down...) Where in that graph are there gaps in my understanding? What are my specific goals and interests?
Can we present the most relevant learning materials to me in alignment with all these parameters? Or else, can we generate them on the fly? It's all a solvable engineering problem.
---
[0] https://nintil.com/bloom-sigma/
[1] Math Problems with GPT-4o [video]: https://news.ycombinator.com/item?id=40347320
[1] Math Problems with GPT-4o: https://www.youtube.com/watch?v=_nSmkyDNulk
See also: One study predicted human-level AI tutors by 2025
[2] Achieving Bloom's Two-Sigma Goal Using Intelligent Tutoring Systems (2020) - https://www.researchgate.net/publication/339246677_Achieving...
Picture this: You're a seventh-grade student who struggles to solve the equations in math. But now you have an AI that solves it for you.
I bet that there will be more students of the later type than the former.
Not that Sal Khan is infallible but he is a good-egg. You're gonna have to give me more than, "this'll never work because of <cynical dime-a-dozen take>."
Sal Khan’s work changed my life and allowed me to build a foundation in not only math, but also finance and economics, that allowed me to feel confident enough to go to college. I did not have a support system or people in my life who could teach me these things, but I had the internet and patience, which meant I had opportunities that did not exist even a few years before (this was in 2014).
Now I work in AI and generally have a life I could not have dreamed of. Every time I go to the grocery store to buy something and don’t have to worry about my card being declined, it feels like magic. Even years later.
If I had been in that situation a few years earlier, my life would be very different.
I’m excited to live in a world where my daughters and countless others will have a tutor that can help them maximize their potential throughout their lives. I’m excited not only for them, but for society. What a time to be alive.