I'd also STRONGLY recommend Polya's "How To Solve It"[0]
It really is about teaching you how to solve arbitrary problems. It teaches how to think. You can read pages xvi and xvii (in [0]) which will give you a good sense at what the book teaches. Half the first chapter is available too (this is an easy book to find online).
It's also technically a math book, but treating it as a standard math book is missing the point (honestly, missing the point of math altogether). It is really about how to solve problems, which that's what math is about. Math is really a collection of logics (plurality intended) that can be used to solve problems. Thing is, this is usually done in the abstract as mathematicians are usually interested in solving the general form of a problem. (The other part being that math is a high precision language, which is another thing it shares with programming languages). People might be scared off by the math notation in it, but I'd wager that anyone who can write anything more complicated than a hello world program is smart enough to make it through this book[1].
You do not need to be a mathematician to benefit from it, nor need to know much math at all. I'll make the strong claim that every single programmer, and anyone doing anything related to STE(A)M will benefit. If the "A" is surprising, I think you'll understand after reading a bit.
[1] Variables in math are no different than in programming. The main difference is math is terse because it's typically hand written. I'd add that programming also has types but most people are writing with languages that do duck typing.
Thank you ever so much for this recommendation. I'm already through half of his upskilling document. It is as if he wrote the book exactly for me. I've been studying first thing in the morning off and on for a while now, and I'm happy to find that he advocates for that. Other than that, his focus on a) maths, b) coding and, c) domain expertise fit in line with where I want to head. I'm also happy to see that he has written much on his blog that I could dig into later.
The importance of "learning how to learn" has been emphasized by all of my teachers since I was in highschool, or maybe even 8th grade, decades ago.
My computer engineering professors also emphasized user centered design. For one of Google's top scientists to bring this up is an admission that they won't, or can't, design a good user experience for their tools.
I remember the same thing. That doesn't mean they knew how to teach us to "learn how to learn". Neither does it mean that the underlying education system supported that goal.
Same goes for user-centered design. Trying to make something user-friendly is one thing, successfully doing it is another. Large organizations are especially poor at user-friendly design because the underlying structures which support that goal don't exist. Organizational science is still in its infancy.
Not to mention successfully measuring success or failure. The overwhelming majority of people I know (from your average Joe to your tenured professor at elite universities, from philosophers to physicists[0]) underestimate the difficulty of measuring things.
Almost everyone treats any metric as they would a ruler or tape measure. Even your standard ruler is not as good of a measuring device as you probably think! But this becomes a huge mess when we start talking about any measurement of statistics or some other abstraction. People treat metrics and algorithms as black boxes, rather than tools. Tools still require craftsmen, who understands: when they work, when they don't work, when they can be used in a pickle, what can be substituted in a pickle, their limits, what new problems they create, and so on. It is incredible how much complexity there is to things that appear so simple. But then again, that's why you get things like an engineering manual on o-rings that is over a thousand pages. And even those aren't comprehensive.
I'm not suggesting we all need to be "master craftsmen", but I actually think we would all do better if we recognized that everything has more depth than might appear. If only to give people a moment of pause to question if they are actually doing things the right way. There's always a better way. The real trick is learning what's good enough and you'll never know what is good enough when everything is simple.
[0] The exception tends to be those that need to work with high precision, since with these jobs you tend to be forced to deal with this in an explicit manner. So more common among people like machinists or experimental physicists. Though sometimes this ends up worse as they can end up operating on vibes. I think it happens when intuitions are successful for too long and not enough meta-analysis is done to update them.
The US education system only has one mode, and thats to survive in a slim way with overworked staff and huge classrooms. 40 kids in a math class is seen as normal.
Everything you see of its character, including emphasizing tests and practice, follows from that. Talking about good UX is miles away.
It's a problem that goes beyond the United States, overworked staff, and constraints in general, although these are legitimate concerns.
I studied in a non-US country, but the attention paid by teachers to pedagogy was virtually zero.
I mean, we had five years of English classes in high school, and by the end of high school, less than five out of 30 people in my cohort were able to string a couple of sentences together in English. And my class was made up of serious, studious young people. It seems to me that the time was not well spent, but did the teacher, a caring and generally competent person, reflect on the poor results? I highly doubt it.
Most teachers want to do better, but are stuck in a system where they're not able to. Overwhelmed with large classes, small budgets, ridgid programs, demanding parents, it's hard to also dedicate energy to reflection or student attention.
Most teachers have very little clue about pedagogy; let's be honest about it. And it does not mean they are incompetent at their subjects or lazy.
The fact that, outside of the expected exceptions, a skill/subject/section is never brought up after the test means that teachers are not thinking at all about retrieval practice.
There is a time for understanding and justifications, and a time for saying things as they are.
Then, my personal experience and that of my friends, who all attended mid- to high-ranking schools outside the United States, is that regardless of class size and teacher workload, teachers never seemed to know how to teach effectively and efficiently.
I found that emphasis to be similar to two teams building from opposite starting points, and never meeting in the middle.
The issue with "learning to learn" is that it does not include the foundational skill of "how to communicate". Far too often it is not a lack of desire to learn, it is the inability to communicate what one is trying to learn. When seeking help, not only does the seeker have difficulty expressing their situation, those trying to help are not taught how to listen and will offer solutions to an issue only starting to be explained. This difficulty is then compounded by self conversation bias that negatively spins against the person seeking. That is two very high hurdles: negative self bias, and inarticulate communications while seeking guidance.
“You have to learn how to learn” has been a phrase often repeated by teachers, but I don’t remember any of them emphasizing, for example, retrieval practice: you learn a skill or subject, move on to the next one, and leave it up to fate whether you remember anything from the first one.
It always surprises and saddens me that, despite having been an excellent student throughout my years of education, I remember practically nothing about 90% of the subjects I studied.
There is a certain amount of "use it or lose it" seemingly inherent to virtually every human endeavor. But I suspect if you were to enroll in a class in any of those subjects, you'd perform radically better than a peer who'd never studied them. IOW, there's often more latent memory than we realize or can easily retrieve.
"IOW, there's often more latent memory than we realize or can easily retrieve." -
I did not find it to be true almost at all, and I tested many other people on it. When I voiced my concerns, the usual answer was, "Yes, but when you pick up a book, you will remember". And then I asked, "Try it", and the subsequent answer was, "I have to admit you are right".
The "re-absorption speed" is heavily confounded by general IQ and the kind of cognitive stimulation one receives in daily life, but the original learnings are mostly gone.
Among other things, this is why retrieval practice is important: it slows down the "forgetting rate".
> Technology is continually reshaping industries and while many eschew learning and adopting, those who embrace it are the ones who succeed best IME.
I said it already in a reply to GP, but I'm going to say it again: I stopped caring about what people list on their resumes, your work history and education don't matter to me. I'd rather hire a hungry junior that finished a bootcamp, that has a drive and ability to absorb new things and adapt to changing environments, over somebody who's got 10 years of experience and can't do shit outside of their comfort zone.
The number of people who aren't able to learn and adapt to changing times, new tools, new ways of working, etc. is shocking.
> that they won't, or can't, design a good user experience for their tools.
This reminds me of old videogames. Many didn't have tutorials. Or rather, they did, but they were the first level. Unlike most modern games many of them would just drop you in and you'd need to figure it out or read the instruction booklet. It definitely helped that there was a common language, that's still mostly used today, but the point was more to let users "discover" the controls themselves. Like here's the start to Wolfenstein 3D[0]. No popup messages, no nothing. A lot of this was done for space savings, but it also forced makers to design in a way that teaches the mechanics and how the world works. You can even see here how the level introduces players to secret rooms. Leading to many players doing the same thing they do in Zelda games, stabbing the walls to see if something is different or pay attention to subtle clues that a secret is here.
It's pretty hard to do this design but I think anyone who's played these games will both admit it is frustrating but rewarding. I think that's true for any learning. The advantage with a videogame is you can provide nearly immediate feedback as well as design feedback delays. I'm highly educated in both math and CS and I think that's actually one of the key differences. When programming there's quick feedback loops. Your program runs or doesn't[1]. Whereas in math you finish a proof and aren't even sure if it is right or not. This does end up teaching different and useful skills, but it sure does create a higher barrier to entry (barrier isn't intelligence, it is persistence).
I think my main concern today (having taught hundreds of college students over the last 5 years) is a lowering in this resilience. I mean I feel it in myself too. We've definitely generated a world where we have quick feedback mechanisms, yet this is impossible to create in more advanced education. It can take weeks, months, or even years to see the real fruits of your labors. I found that in classes where we had autograders or provided students with test cases[2] that often these ended up hurting the students more than helping. They became over-reliant on them, outsourcing their thinking to what we were providing as aids. I watched ChatGPT come out during this time and was not surprised that this only furthered the problem. I was only a grad student, so most classes I did not have good control and sometimes not much of a say, but if I were to do it again I'd try to push the aids out more slowly[3]. The most common problem was that students wrote to the test, not to the requirements. It's actually not a uncommon outside school, and I see a lot of people do quite similar things in industry. Thinking that passing tests is sufficient. But writing to tests will only result in you being as complete as the tests. It's a failed paradigm, you'll never have full coverage.[4]
There's definitely other problems with the education system and I don't want to dismiss them. There's no cureall, but I think this might be something most people might want to think about. Despite saying these words, they are still something I need to reinforce. Good habits are hard to maintain and it is only becoming easier to unknowingly slip into bad ones.
[1] Well there's the secret third option which is the most common: it runs, but doesn't run like you think it runs.
[2] We always stated that these were incomplete, that they should write additional ones. There probably wasn't a single office hour I held where I hadn't mentioned that one can never have complete coverage through tests.
[3] I don't know the full answer but here's something I would try. We had autograders and allowed students to submit as many times as they wanted. I'd keep these, but re-implement to have an exponential backoff (until say 2hrs before the assignment was due. I honestly never marked anyone late unless it hadn't been submitted by noon the next day). After a few failed submissions, pass them a subset of test cases. Then repeat. This definitely puts a lot more work on our end as the educator, but it would put students into a position where they need think about the problem first. That's a critical self-learning strategy. The struggle is necessary for success. Too often people just want to jump to the end, assuming a well defined answer already exists. They'll find something that looks appropriate, implement it, and declare success while missing the devil hiding in the details. Too early of feedback only reinforces that strategy.
[4] I'm certain someone will read this believing I am suggesting no tests. I assure you, if this was your interpretation then your interpretation is wrong. Trust me, it is my thoughts I'm trying to convey.
Thing is, this has nothing to do with UX or even AI at the end of the day. Over the years I have adjusted the way I handle interviews when I'm on a hiring panel to focus on critical thinking, problem solving, and lifelong learning above all else - because they are the #1 indicator that a hire is going to be successful in a position, even if that means they need a lot more training before they become productive.
I can teach somebody who finished a 6-month coding bootcamp Go, all the internal tooling, go over the business with them, etc. if they have these skills and end up with a productive mid-level engineer who gets shit done in a few years. What I can't teach is the drive and ability to learn, that's a much longer process and if you don't already have it then I'm not prepared to develop it.
Hell, outside of looking for signs of obvious bullshit I stopped giving a shit about resumes. Your work history does me no good, your education doesn't matter to me, and your references are useless beyond making sure you aren't straight up lying to me about your employment history. Every single time I have hired somebody who has 5 years of "experience" working with technologies I bullet pointed on a JD they ended up fumbling the moment they had to do something new. Doing leetcode, pair programming sessions, take-home assignments, whiteboarding system designs, etc. for SWE positions did nothing to really improve this; for SRE/DevOps roles I tried trivia questions (how are containers implemented - like what kernel technologies do they use and what do they do, how would you go about investigating why a service is consuming 100% CPU time), throwing them at broken VM's and more take-home assignments.
AI tools only make this skillset more important - I can throw Junie, Claude Code, or Copilot and small task and end up with...an implementation. But they still fuck up, constantly, and yet again, anything that's not already been done, regularly, requires a lot of guidance from an engineer in the loop. And with the god damned death of the web thanks to AI slop being posted anywhere, the ability to find answers and reason through problems is only going to become more important when these tools fail miserably for the third time in a row.
> Muad'Dib learned rapidly because his first training was in how to learn. And the first lesson of all was the basic trust that he could learn. It is shocking to find how many people do not believe they can learn, and how many more believe learning to be difficult. Muad'Dib knew that every experience carries a lesson.
-- from "The Humanity of Muad'Dib" by the Princess Irulan
That story had a profound impact on me when I read it as a kid. It encouraged me to learn for the sake of learning and to not get worked up over some test result. Many times the test is measuring something different than you think, than even the testers think. It encouraged me to look at things differently, so find more depth. To realize that to make things different you also have to do things differently.
Reading again as I've become older I can see Asimov's call to fix education in the US and how much he resented the growing anti-scientific movements. As a child I thought it would be cool to belong to this elite group of "tape makers" (people who create knowledge)[0]. But as an adult I realized the naivety and grotesqueness of a society constructed in such a way. It is inhumane because it turns man into machine, with only a few being allowed to even be given the chance to explore the world around them.
It's a really good read. I highly recommend. It's old Sci-Fi, so can be rough, but I feel it has aged well. It's under 50 pages, so easy to finish in a single sitting. It's in Asimov's "The Complete Stories Vol I" if you find a copy, but you can find it online or I believe it is also in "The Asimov Chronicles". It's also HN, so I hope everyone knows ways to get those...
Btw, the Wiki[1] has spoilers. Do not read the paragraph starting with "Suddenly a stranger appears" nor the one that follows. Here's a short non-spoiler version instead: In the 66th century, school doesn't exist and instead people are educated through a direct download into their brain. They learn to read at 8 and are fully educated at 18, allowing them to live their lives as children do. Our protagonist, George, differs than his peers, being a nerd and using his new found ability to read to read whatever he can get his hands on. At 18, children are fully educated, tested, and then assigned careers by their aptitudes. George stands out, his test results suggest he is mentally challenged, is not given a designated career, and thus is not allowed to be further educated. Not being fit for any career or education, they commit him to psychiatric ward where he'll live out the rest of his days.
[0] In some sense I still do. I have a deep passion for research and my dream job would be somebody giving me money to just explore my ideas. Probably not too far off from what people imagine a tenured professor does, though that's quite different from what one actually does.
> “Not infinite. Architects. Teachers. Teachers of teachers, but the art of teaching teaching is much the same as the art of teaching. Three levels is enough. Though the levels have to mix. The teacher who trains the next architect must be a master both of teaching and of architecture. I will spare you the math, but one needs a series of teachers at different points on the teaching-skill/architecture-skill tradeoff-curve. One will be a master teacher who has devoted decades to learning the textbook-writing skill, and who can write a brilliant Introduction To Architecture textbook that makes the first ten years of architecture ability seem perfectly natural and easy to master. Another will be a mediocre teacher who knows enough advanced architecture to write a passable textbook on the subject. Still another will do nothing but study pure Teaching itself, in the hopes that he can one day pass on this knowledge to others who will use it to write architecture textbooks.
One issue that is not discussed enough when talking about learning is mental preparation for learning. We have all had days when learning seemed easier than on other days, but we didn't pay too much attention to it, or we thought that the subject we were learning was more favorable to us, or we classified it as one of the many inexplicable or unrepeatable circumstances of life.
While we understand the importance of warming up for physical activity and recognize the need for a certain aptitude for running, weightlifting, or boxing, when it comes to more intellectual activities, we often leave things to chance: sometimes we are more alert and receptive, while at other times we are less so.
Over the years, I have found enormous benefit in practicing autogenic training, a more Western and scientific version of meditative practices that today seem to arouse the interest of those who deal with these things. I am mentally more alert, more receptive, and learning, which is always challenging, is faster.
> We have all had days when learning seemed easier than on other days
Unfortunately it's hard to trust our feelings on this. There is a lot of literature that demonstrates an inverse relationship between how well we felt we learned and how well we actually learned.
> mental preparation for learning
This is actually really big in the learning literature. It's not meditation, though, it's priming through pretesting and prior knowledge activation. These types of "warm ups" have outsized effects on retention and understanding.
> autogenic training, a more Western and scientific version of meditative practices
Do you have any tips for learning more and getting started? I have searched a bit, but always appreciate anecdotes of those that have found success enough to speak about it.
I will probably write a book on this topic because there is not enough material in English. But the practice is fairly simple overall and only requires time and a positive attitude. Since there is a somatic response, it is always easy to tell if it is “working", especially for the "lower" exercises.
Originally, it had to be taught by MDs, according to Dr. Schultz--the inventor of the method--and his followers, but that ship has long sailed.
It is fairly easy to find a copy of Karl Rosa's book “Autogenic Training”, a good starting point, and Luthe's in-depth multi-volume analysis of autogenic training, though I would only recommend the latter to the most avid enthusiasts.
If you need to learn how to learn then you don't know how to learn. How can you learn to learn if you don't know how to learn?
Jokes aside I'm really into learning science and make youtube videos covering learning and learning papers + an ipad app. I keep a running list of my favorite learn-to-learn resources here:
If I had to recommend only one resource it would be: The ABCs of How We Learn: 26 Scientifically Proven Approaches, How They Work, and When to Use Them by Schwartz
Even though you might be joking, I think this is a very relevant point. To take it a step further, one also has to be willing to learn to learn how to learn. I wonder if intrinsic motivation for this can be nurtured, or whether it is up to chance.
I think the learning science itself offers solutions to this. Intrinsic motivation is apart of learning theory, specifically in the affective domain.
In practice, the problem you run into, even with intrinsically motivated learners, is that they will not use the active learning techniques they are studying. Often times they will revert back to rote memorization, highlighting, re-reading, copying notes, cramming, etc rather than use the things they are learning about learning to learn the subject of learning itself.
I think you have to start with:
1) Teaching learning in the first place. No one teaches learning how to learn, so we should just start there. We already have captive audiences in the form of schools, we just put the cart before the horse and teach subjects and hope the skill of learning emerges. This is poor pedagogy.
2) Work with the students and ensure that they are actually using the learning techniques they are being taught on the subject of learning itself. This is the only way I've seen it work.
If you try to learn how to learn using passive learning techniques, you won't learn the subject of learning, which is what I think OP was referring to. People who do not know how to learn use passive learning techniques which results in rapid forgetting. They have to use the active learning techniques they are learning on the subject of learning itself.
There are so many barriers to good learning. I like the saying "Maslow before Bloom", which refers to the two popular taxonomies. It succinctly captures the idea that students need food, shelter, and safety in order to learn effectively.
I know we're all good little rational kids here, but even rationalists need to learn about emotions. Strong emotional responses are currently holding back human advancement. If you look closely at history, it has been always thus.
** "in my opinion" is always implied, unless a source is given **
Reading about airline crashes has radically changed how I view blame.
The way I was raised and the choices I made as an adult have given me a relatively rare point of view: people are made of humans, and humans are made of animals, and animals have limited capabilities.
I can explain someone's actions, or I can excuse someone's actions, and the difference is largely in the mind of the beholder.
Social punishment is micro and macro. On the macro it looks like shared morality and it feels like safety. On the micro it looks like emotional invalidation and it feels like danger and isolation.
I run a program for high schoolers to emphasize this skill. However, the entire K-University pipeline is designed around credentialism. Ie. do whatever you need to, cram/cheat/regurgitate, to get the rubber stamp. It's really hard to communicate the importance of self-directed education/learning how to learn when the vast majority of students' formal educational experiences tell them otherwise. Very frustrating but perhaps things are changing ...
School has two competing goals and this will never change:
1. Have the kids learn new things
2. Have the kids reach a desired level of competency
Learning happens where you are at, not where the teacher wants you to be. Every student is at a different place in understanding. It's impossible without 1-on-1 instruction to really maximize learning.
Competency is only determined via testing. Learning doesn't require testing at all, you can just speak to a student to get a good idea if they're making some progress, any progress. Competency? That basically demands a test, because it has a particular bar in mind.
Now students know they need to pass the bar, somehow, but the anxiety of that is going to cause issues with them just trying to learn. This is unfixable though, because the outside pressures demand students have some level of competency otherwise teachers are viewed as failures.
I agree. Imo, #2 is becoming more of an emphasis over time. Teacher don't have much time/energy to pursue #1. Eventually, most of them stop caring and rely on testing metrics because that's what the admins want.
It's amazing what kids can learn if they just spent a little bit of time with a 1-on-1 instructor/advisor. The anxiety you mentioned can be crippling and something I deal with regularly. Even some of the "gifted" kids (perhaps due to the expectations) have trouble avoiding the trap of overindexing on productivity/competency metrics. They're not even self aware of it, just accepts it as normal.
For most kids I have to go through the exercise of separating these two concerns, the learning part and the signaling part, early so they can put things in perspective.
There's also the validity of learning methods, despite what studies may claim, there's no scientific "grand theory of meta-learning", and if ideas are misapplied/misused there's a risk of falling into scientism, which would be just as harmful as economically driven credentialism. At worst it is just the austerity version of education—learn it yourself because we can't afford the school resources to teach/coach/nurture subjects.
There is no grand theory perhaps, but educational research is not totally useless, people have accumulated a lot of insights over the years and some of these "intuitions" are also backed by studies.
Here are some book-length reviews of currently known things about learning/teaching that I found to have a very high signal-to-noise ratio:
We’re in dire need of this right now. The number of people that I work with who refuse to pick up new tools and technologies is astounding. If they _do_ try something new, they seem to avoid all but the most basic knowledge of whatever it is, and look at me crosseyed if I suggest going the slightest bit deeper (`git add -p` rather than `git add .`, for example).
I'm sure it varies a lot place to place but I've experienced much more of the opposite in the tech industry. I've heard countless times that we should switch to a different tool because it's newer, from someone who couldn't name a single specific way it's better than the existing tool. I see so much busy work at my employers and in products I use where things get changed just for the sake of change, without getting any better.
I feel like it's a mix of both, depending on how familiar thungs are. It's probably much easier for someone in tech to try moving to a new tool for no real reason than it is for them to learn a bit of CAD and a bit of electronics to make a widget (again, for no real reason).
I think that's a somewhat unfair framing of the issue, since it isn't a lack of effort from teachers, but rather the pressures from higher levels towards increasing test scores. LLMs don't really change that.
There's also the fact that kids are being taught the very basics, the sorts of things increasingly intelligent models are most likely to be able to solve first. I don't think there's any level of effort that can be put into designing assignments to get around this.
Similar to how teachers haven't really been able to do anything to stop kids from sticking their algebra problems into wolframalpha or other tools besides just making them do the work in class (which then cuts into teaching time).
Beyond a certain point, all that can be done is for teachers to try to instill the importance of practice into students, and for parents to be more proactive in monitoring how their kid is doing their homework.
Maybe we'll see an increase in after-school classes for kids to do homework in, under teacher supervision.
> If we're lucky, LLMs force people to put more effort into assignments and grading and then that would help kids learn to learn as well.
I'm afraid it might be exactly opposite. Having all the knowledge at hand. all the time will lead to knowledge atrophy. Just like it already happens with ability to travel without navigation.
With how rapidly LLMs are improving, I don't know how you would construct assignments that can't be solved relatively quickly by a student feeding it into a bleeding-edge LLM. Especially since teachers often aren't PhDs and are overworked, the idea of every class of students getting handed brand new problems that aren't in the training set feels far-fetched.
I hope somebody figures this out but I don't know what the solution looks like.
first off it doesn't seem to be taught at the moment, but also I'm pretty sure that has always been the most important and foundational skill, and it seems like there might be an upper bound for what percentage of people can actually learn it.
Learning is a nontrivial skill even though it has historically been treated as such. It requires an embodied understanding of concepts from basic cognitive psychology, expertise theory, behavioral-affective psychology, metacognition, and more. Until people stop with the platitudes of "learning how to learn is important" and start teaching/training the subject directly as a skill that must be acquired, this will not change.
Simply showing a learner a few slides on spaced retrieval will not cut it.
I'd place this skill, "Learning how to Learn", with Cal Newport's notion of "Deep Work". Part of me wants to say that the latter is a precondition of the former, but I'm not sure that's the case.
A lot of what people think of as "doing stuff" relies on years or decades of training and experience. When you pay someone to maintain part of your house or to create bespoke furniture or repair your car you're not just paying for labor, you're paying for labor from someone with knowledge and equipment.
I think there's an element of curiosity that is innate, and that make a huge difference. For those so blessed, learning is fun and a welcome activity (there's an entirely separate issue of actually applying any of knowledge but at least it's a start).
That said, teachers can make all the difference in connecting with students to guide them into discovery vs. cramming facts into their heads to satisfy testing metrics.
Terrible article. Did Hassabis really say ”learning how to learn” is itself a skill, like the article paraphrases it? Surely the skill isn’t that you can learn how to learn, the skill is that you know how to learn. Just like ”learning to ride a bike” isn’t a skill; it’s something you do once, leading to a useful skill.
”Learning how to learn” sounds vaguely insightful just because of the repetition, but if you think for a bit about what it actually means it falls apart.
As a Curriculum Designer and OG dial-up Millennial I agree wholeheartedly. Too late though. Enlightenment, collaboration, and advancing human experience isn't profitable.
In other words, until one learns how to hammer a nail, it's unreasonable to assume knowledge of how to tell another to do so. AI is no exception. It's speed-running US society's final threads being severed, and okay, sigh, here we go. No, I'm not interested in fixing the problems he's identifying.
My ex had a saying from bench science..."if you're not part of the solution, you're part of the precipitant." That part. Off to go live in a van down by the river...
Join the Generalist bootcamp, it includes big picture of the world and everything, anything you ever need. Full access subscriptions at $1000.
You will Learn the following things:
Analytic philosophy,
Mathematical logic,
Pure and applied math,
Physics, CS, Systems thinking,
Engineering(Mech + electronics),
creative problem solving
And finally one art subject
Beginer Projects:
Wafer stage design.
Model nano tech projects.
Small nuclear fusion reactors.
Portable TEM machine.
Pre-req:
Just enough maturity.
You should be curious, persistence & hardworking.
We assume you will practise problem solving till you die.
Outcomes of the bootcamp:
Job guaranteed at fortune 500.
Testimonials:
We have so many happy customers working for companies having trillion dollar values.
For decades, I've said
You don't go to college to get a job
You don't go to college to get a degree
You don't go to college to learn things
You go to college to LEARN HOW TO LEARN
If you do that, all the other things will come to you.
If you get out of college and have not learned how to learn, your "degree" is toilet paper.
The best practical book I found that is way better than overly generic books on "deep work" is Jason Skycak's "Advice on Upskilling"
https://www.justinmath.com/books/ (scroll for the second one)
He is a math guy, who worked in Wall Street and then left and now works on math academy buidling models to improve learning.
It's great, well researched and practical book. However, it's not easy at all. Go check it out. It's free and he has published google docs version.
I'd also STRONGLY recommend Polya's "How To Solve It"[0]
It really is about teaching you how to solve arbitrary problems. It teaches how to think. You can read pages xvi and xvii (in [0]) which will give you a good sense at what the book teaches. Half the first chapter is available too (this is an easy book to find online).
It's also technically a math book, but treating it as a standard math book is missing the point (honestly, missing the point of math altogether). It is really about how to solve problems, which that's what math is about. Math is really a collection of logics (plurality intended) that can be used to solve problems. Thing is, this is usually done in the abstract as mathematicians are usually interested in solving the general form of a problem. (The other part being that math is a high precision language, which is another thing it shares with programming languages). People might be scared off by the math notation in it, but I'd wager that anyone who can write anything more complicated than a hello world program is smart enough to make it through this book[1].
You do not need to be a mathematician to benefit from it, nor need to know much math at all. I'll make the strong claim that every single programmer, and anyone doing anything related to STE(A)M will benefit. If the "A" is surprising, I think you'll understand after reading a bit.
[0] https://www.amazon.com/How-Solve-Mathematical-Princeton-Scie...
[1] Variables in math are no different than in programming. The main difference is math is terse because it's typically hand written. I'd add that programming also has types but most people are writing with languages that do duck typing.
Thank you ever so much for this recommendation. I'm already through half of his upskilling document. It is as if he wrote the book exactly for me. I've been studying first thing in the morning off and on for a while now, and I'm happy to find that he advocates for that. Other than that, his focus on a) maths, b) coding and, c) domain expertise fit in line with where I want to head. I'm also happy to see that he has written much on his blog that I could dig into later.
I looked through the google docs version 0 on my phone, it’s pretty interesting. Is there an epub or PDFs versions?
0 https://docs.google.com/document/u/0/d/17qFY5w9uEWL4VVyJSTEm...
You can export the google docs version to epub:
File > Download > epub
The importance of "learning how to learn" has been emphasized by all of my teachers since I was in highschool, or maybe even 8th grade, decades ago.
My computer engineering professors also emphasized user centered design. For one of Google's top scientists to bring this up is an admission that they won't, or can't, design a good user experience for their tools.
I remember the same thing. That doesn't mean they knew how to teach us to "learn how to learn". Neither does it mean that the underlying education system supported that goal.
Same goes for user-centered design. Trying to make something user-friendly is one thing, successfully doing it is another. Large organizations are especially poor at user-friendly design because the underlying structures which support that goal don't exist. Organizational science is still in its infancy.
Almost everyone treats any metric as they would a ruler or tape measure. Even your standard ruler is not as good of a measuring device as you probably think! But this becomes a huge mess when we start talking about any measurement of statistics or some other abstraction. People treat metrics and algorithms as black boxes, rather than tools. Tools still require craftsmen, who understands: when they work, when they don't work, when they can be used in a pickle, what can be substituted in a pickle, their limits, what new problems they create, and so on. It is incredible how much complexity there is to things that appear so simple. But then again, that's why you get things like an engineering manual on o-rings that is over a thousand pages. And even those aren't comprehensive.
I'm not suggesting we all need to be "master craftsmen", but I actually think we would all do better if we recognized that everything has more depth than might appear. If only to give people a moment of pause to question if they are actually doing things the right way. There's always a better way. The real trick is learning what's good enough and you'll never know what is good enough when everything is simple.
[0] The exception tends to be those that need to work with high precision, since with these jobs you tend to be forced to deal with this in an explicit manner. So more common among people like machinists or experimental physicists. Though sometimes this ends up worse as they can end up operating on vibes. I think it happens when intuitions are successful for too long and not enough meta-analysis is done to update them.
The US education system only has one mode, and thats to survive in a slim way with overworked staff and huge classrooms. 40 kids in a math class is seen as normal.
Everything you see of its character, including emphasizing tests and practice, follows from that. Talking about good UX is miles away.
It's a problem that goes beyond the United States, overworked staff, and constraints in general, although these are legitimate concerns. I studied in a non-US country, but the attention paid by teachers to pedagogy was virtually zero.
I mean, we had five years of English classes in high school, and by the end of high school, less than five out of 30 people in my cohort were able to string a couple of sentences together in English. And my class was made up of serious, studious young people. It seems to me that the time was not well spent, but did the teacher, a caring and generally competent person, reflect on the poor results? I highly doubt it.
Most teachers want to do better, but are stuck in a system where they're not able to. Overwhelmed with large classes, small budgets, ridgid programs, demanding parents, it's hard to also dedicate energy to reflection or student attention.
Most teachers have very little clue about pedagogy; let's be honest about it. And it does not mean they are incompetent at their subjects or lazy.
The fact that, outside of the expected exceptions, a skill/subject/section is never brought up after the test means that teachers are not thinking at all about retrieval practice. There is a time for understanding and justifications, and a time for saying things as they are.
Then, my personal experience and that of my friends, who all attended mid- to high-ranking schools outside the United States, is that regardless of class size and teacher workload, teachers never seemed to know how to teach effectively and efficiently.
I found that emphasis to be similar to two teams building from opposite starting points, and never meeting in the middle.
The issue with "learning to learn" is that it does not include the foundational skill of "how to communicate". Far too often it is not a lack of desire to learn, it is the inability to communicate what one is trying to learn. When seeking help, not only does the seeker have difficulty expressing their situation, those trying to help are not taught how to listen and will offer solutions to an issue only starting to be explained. This difficulty is then compounded by self conversation bias that negatively spins against the person seeking. That is two very high hurdles: negative self bias, and inarticulate communications while seeking guidance.
“You have to learn how to learn” has been a phrase often repeated by teachers, but I don’t remember any of them emphasizing, for example, retrieval practice: you learn a skill or subject, move on to the next one, and leave it up to fate whether you remember anything from the first one.
It always surprises and saddens me that, despite having been an excellent student throughout my years of education, I remember practically nothing about 90% of the subjects I studied.
There is a certain amount of "use it or lose it" seemingly inherent to virtually every human endeavor. But I suspect if you were to enroll in a class in any of those subjects, you'd perform radically better than a peer who'd never studied them. IOW, there's often more latent memory than we realize or can easily retrieve.
"IOW, there's often more latent memory than we realize or can easily retrieve." -
I did not find it to be true almost at all, and I tested many other people on it. When I voiced my concerns, the usual answer was, "Yes, but when you pick up a book, you will remember". And then I asked, "Try it", and the subsequent answer was, "I have to admit you are right".
The "re-absorption speed" is heavily confounded by general IQ and the kind of cognitive stimulation one receives in daily life, but the original learnings are mostly gone. Among other things, this is why retrieval practice is important: it slows down the "forgetting rate".
> One thing we'll know for sure is you're going to have to continually learn ... throughout your career," he said.
This has been the case for literally my entire career and I assume most of the professional world for the last half century.
Technology is continually reshaping industries and while many eschew learning and adopting, those who embrace it are the ones who succeed best IME.
> Technology is continually reshaping industries and while many eschew learning and adopting, those who embrace it are the ones who succeed best IME.
I said it already in a reply to GP, but I'm going to say it again: I stopped caring about what people list on their resumes, your work history and education don't matter to me. I'd rather hire a hungry junior that finished a bootcamp, that has a drive and ability to absorb new things and adapt to changing environments, over somebody who's got 10 years of experience and can't do shit outside of their comfort zone.
The number of people who aren't able to learn and adapt to changing times, new tools, new ways of working, etc. is shocking.
It's pretty hard to do this design but I think anyone who's played these games will both admit it is frustrating but rewarding. I think that's true for any learning. The advantage with a videogame is you can provide nearly immediate feedback as well as design feedback delays. I'm highly educated in both math and CS and I think that's actually one of the key differences. When programming there's quick feedback loops. Your program runs or doesn't[1]. Whereas in math you finish a proof and aren't even sure if it is right or not. This does end up teaching different and useful skills, but it sure does create a higher barrier to entry (barrier isn't intelligence, it is persistence).
I think my main concern today (having taught hundreds of college students over the last 5 years) is a lowering in this resilience. I mean I feel it in myself too. We've definitely generated a world where we have quick feedback mechanisms, yet this is impossible to create in more advanced education. It can take weeks, months, or even years to see the real fruits of your labors. I found that in classes where we had autograders or provided students with test cases[2] that often these ended up hurting the students more than helping. They became over-reliant on them, outsourcing their thinking to what we were providing as aids. I watched ChatGPT come out during this time and was not surprised that this only furthered the problem. I was only a grad student, so most classes I did not have good control and sometimes not much of a say, but if I were to do it again I'd try to push the aids out more slowly[3]. The most common problem was that students wrote to the test, not to the requirements. It's actually not a uncommon outside school, and I see a lot of people do quite similar things in industry. Thinking that passing tests is sufficient. But writing to tests will only result in you being as complete as the tests. It's a failed paradigm, you'll never have full coverage.[4]
There's definitely other problems with the education system and I don't want to dismiss them. There's no cureall, but I think this might be something most people might want to think about. Despite saying these words, they are still something I need to reinforce. Good habits are hard to maintain and it is only becoming easier to unknowingly slip into bad ones.
[0] https://www.youtube.com/watch?v=n0esIfiOGFA
[1] Well there's the secret third option which is the most common: it runs, but doesn't run like you think it runs.
[2] We always stated that these were incomplete, that they should write additional ones. There probably wasn't a single office hour I held where I hadn't mentioned that one can never have complete coverage through tests.
[3] I don't know the full answer but here's something I would try. We had autograders and allowed students to submit as many times as they wanted. I'd keep these, but re-implement to have an exponential backoff (until say 2hrs before the assignment was due. I honestly never marked anyone late unless it hadn't been submitted by noon the next day). After a few failed submissions, pass them a subset of test cases. Then repeat. This definitely puts a lot more work on our end as the educator, but it would put students into a position where they need think about the problem first. That's a critical self-learning strategy. The struggle is necessary for success. Too often people just want to jump to the end, assuming a well defined answer already exists. They'll find something that looks appropriate, implement it, and declare success while missing the devil hiding in the details. Too early of feedback only reinforces that strategy.
[4] I'm certain someone will read this believing I am suggesting no tests. I assure you, if this was your interpretation then your interpretation is wrong. Trust me, it is my thoughts I'm trying to convey.
Thing is, this has nothing to do with UX or even AI at the end of the day. Over the years I have adjusted the way I handle interviews when I'm on a hiring panel to focus on critical thinking, problem solving, and lifelong learning above all else - because they are the #1 indicator that a hire is going to be successful in a position, even if that means they need a lot more training before they become productive.
I can teach somebody who finished a 6-month coding bootcamp Go, all the internal tooling, go over the business with them, etc. if they have these skills and end up with a productive mid-level engineer who gets shit done in a few years. What I can't teach is the drive and ability to learn, that's a much longer process and if you don't already have it then I'm not prepared to develop it.
Hell, outside of looking for signs of obvious bullshit I stopped giving a shit about resumes. Your work history does me no good, your education doesn't matter to me, and your references are useless beyond making sure you aren't straight up lying to me about your employment history. Every single time I have hired somebody who has 5 years of "experience" working with technologies I bullet pointed on a JD they ended up fumbling the moment they had to do something new. Doing leetcode, pair programming sessions, take-home assignments, whiteboarding system designs, etc. for SWE positions did nothing to really improve this; for SRE/DevOps roles I tried trivia questions (how are containers implemented - like what kernel technologies do they use and what do they do, how would you go about investigating why a service is consuming 100% CPU time), throwing them at broken VM's and more take-home assignments.
AI tools only make this skillset more important - I can throw Junie, Claude Code, or Copilot and small task and end up with...an implementation. But they still fuck up, constantly, and yet again, anything that's not already been done, regularly, requires a lot of guidance from an engineer in the loop. And with the god damned death of the web thanks to AI slop being posted anywhere, the ability to find answers and reason through problems is only going to become more important when these tools fail miserably for the third time in a row.
> Muad'Dib learned rapidly because his first training was in how to learn. And the first lesson of all was the basic trust that he could learn. It is shocking to find how many people do not believe they can learn, and how many more believe learning to be difficult. Muad'Dib knew that every experience carries a lesson.
-- from "The Humanity of Muad'Dib" by the Princess Irulan
Also, The Profession by Isaak Asimov is entirely dedicated to the importance of learning.
https://classicsofsciencefiction.com/2024/04/05/profession-b...
That story had a profound impact on me when I read it as a kid. It encouraged me to learn for the sake of learning and to not get worked up over some test result. Many times the test is measuring something different than you think, than even the testers think. It encouraged me to look at things differently, so find more depth. To realize that to make things different you also have to do things differently.
Reading again as I've become older I can see Asimov's call to fix education in the US and how much he resented the growing anti-scientific movements. As a child I thought it would be cool to belong to this elite group of "tape makers" (people who create knowledge)[0]. But as an adult I realized the naivety and grotesqueness of a society constructed in such a way. It is inhumane because it turns man into machine, with only a few being allowed to even be given the chance to explore the world around them.
It's a really good read. I highly recommend. It's old Sci-Fi, so can be rough, but I feel it has aged well. It's under 50 pages, so easy to finish in a single sitting. It's in Asimov's "The Complete Stories Vol I" if you find a copy, but you can find it online or I believe it is also in "The Asimov Chronicles". It's also HN, so I hope everyone knows ways to get those...
Btw, the Wiki[1] has spoilers. Do not read the paragraph starting with "Suddenly a stranger appears" nor the one that follows. Here's a short non-spoiler version instead: In the 66th century, school doesn't exist and instead people are educated through a direct download into their brain. They learn to read at 8 and are fully educated at 18, allowing them to live their lives as children do. Our protagonist, George, differs than his peers, being a nerd and using his new found ability to read to read whatever he can get his hands on. At 18, children are fully educated, tested, and then assigned careers by their aptitudes. George stands out, his test results suggest he is mentally challenged, is not given a designated career, and thus is not allowed to be further educated. Not being fit for any career or education, they commit him to psychiatric ward where he'll live out the rest of his days.
[0] In some sense I still do. I have a deep passion for research and my dream job would be somebody giving me money to just explore my ideas. Probably not too far off from what people imagine a tenured professor does, though that's quite different from what one actually does.
[1] https://en.wikipedia.org/wiki/Profession_(novella)
Irulan’s books are state propaganda! The true Paul Atreides is only revealed in Leto II’s secret diaries.
Ars longa vita Brevis by Scott Alexander also has some interesting perspectives on the art of learning and teaching:
https://slatestarcodex.com/2017/11/09/ars-longa-vita-brevis/
> “Not infinite. Architects. Teachers. Teachers of teachers, but the art of teaching teaching is much the same as the art of teaching. Three levels is enough. Though the levels have to mix. The teacher who trains the next architect must be a master both of teaching and of architecture. I will spare you the math, but one needs a series of teachers at different points on the teaching-skill/architecture-skill tradeoff-curve. One will be a master teacher who has devoted decades to learning the textbook-writing skill, and who can write a brilliant Introduction To Architecture textbook that makes the first ten years of architecture ability seem perfectly natural and easy to master. Another will be a mediocre teacher who knows enough advanced architecture to write a passable textbook on the subject. Still another will do nothing but study pure Teaching itself, in the hopes that he can one day pass on this knowledge to others who will use it to write architecture textbooks.
One issue that is not discussed enough when talking about learning is mental preparation for learning. We have all had days when learning seemed easier than on other days, but we didn't pay too much attention to it, or we thought that the subject we were learning was more favorable to us, or we classified it as one of the many inexplicable or unrepeatable circumstances of life.
While we understand the importance of warming up for physical activity and recognize the need for a certain aptitude for running, weightlifting, or boxing, when it comes to more intellectual activities, we often leave things to chance: sometimes we are more alert and receptive, while at other times we are less so.
Over the years, I have found enormous benefit in practicing autogenic training, a more Western and scientific version of meditative practices that today seem to arouse the interest of those who deal with these things. I am mentally more alert, more receptive, and learning, which is always challenging, is faster.
> We have all had days when learning seemed easier than on other days
Unfortunately it's hard to trust our feelings on this. There is a lot of literature that demonstrates an inverse relationship between how well we felt we learned and how well we actually learned.
> mental preparation for learning
This is actually really big in the learning literature. It's not meditation, though, it's priming through pretesting and prior knowledge activation. These types of "warm ups" have outsized effects on retention and understanding.
> autogenic training, a more Western and scientific version of meditative practices
Do you have any tips for learning more and getting started? I have searched a bit, but always appreciate anecdotes of those that have found success enough to speak about it.
I will probably write a book on this topic because there is not enough material in English. But the practice is fairly simple overall and only requires time and a positive attitude. Since there is a somatic response, it is always easy to tell if it is “working", especially for the "lower" exercises.
Originally, it had to be taught by MDs, according to Dr. Schultz--the inventor of the method--and his followers, but that ship has long sailed.
It is fairly easy to find a copy of Karl Rosa's book “Autogenic Training”, a good starting point, and Luthe's in-depth multi-volume analysis of autogenic training, though I would only recommend the latter to the most avid enthusiasts.
If you need to learn how to learn then you don't know how to learn. How can you learn to learn if you don't know how to learn?
Jokes aside I'm really into learning science and make youtube videos covering learning and learning papers + an ipad app. I keep a running list of my favorite learn-to-learn resources here:
https://www.ahmni.app/blog/learn-to-learn-resource-list
If I had to recommend only one resource it would be: The ABCs of How We Learn: 26 Scientifically Proven Approaches, How They Work, and When to Use Them by Schwartz
Even though you might be joking, I think this is a very relevant point. To take it a step further, one also has to be willing to learn to learn how to learn. I wonder if intrinsic motivation for this can be nurtured, or whether it is up to chance.
I think the learning science itself offers solutions to this. Intrinsic motivation is apart of learning theory, specifically in the affective domain.
In practice, the problem you run into, even with intrinsically motivated learners, is that they will not use the active learning techniques they are studying. Often times they will revert back to rote memorization, highlighting, re-reading, copying notes, cramming, etc rather than use the things they are learning about learning to learn the subject of learning itself.
I think you have to start with:
1) Teaching learning in the first place. No one teaches learning how to learn, so we should just start there. We already have captive audiences in the form of schools, we just put the cart before the horse and teach subjects and hope the skill of learning emerges. This is poor pedagogy.
2) Work with the students and ensure that they are actually using the learning techniques they are being taught on the subject of learning itself. This is the only way I've seen it work.
If you try to learn how to learn using passive learning techniques, you won't learn the subject of learning, which is what I think OP was referring to. People who do not know how to learn use passive learning techniques which results in rapid forgetting. They have to use the active learning techniques they are learning on the subject of learning itself.
There are so many barriers to good learning. I like the saying "Maslow before Bloom", which refers to the two popular taxonomies. It succinctly captures the idea that students need food, shelter, and safety in order to learn effectively.
Long has been.
Best bit of career advice I ever got, back in the 90s: "Get really good at the help system".
(At the time, it was MSDN DVDs).
I know we're all good little rational kids here, but even rationalists need to learn about emotions. Strong emotional responses are currently holding back human advancement. If you look closely at history, it has been always thus.
** "in my opinion" is always implied, unless a source is given **
Reading about airline crashes has radically changed how I view blame.
The way I was raised and the choices I made as an adult have given me a relatively rare point of view: people are made of humans, and humans are made of animals, and animals have limited capabilities.
I can explain someone's actions, or I can excuse someone's actions, and the difference is largely in the mind of the beholder.
Social punishment is micro and macro. On the macro it looks like shared morality and it feels like safety. On the micro it looks like emotional invalidation and it feels like danger and isolation.
there is a book/course by that name, it's decent I would say https://www.amazon.com/Learning-How-Learn-Spending-Studying/... https://www.coursera.org/learn/learning-how-to-learn
I run a program for high schoolers to emphasize this skill. However, the entire K-University pipeline is designed around credentialism. Ie. do whatever you need to, cram/cheat/regurgitate, to get the rubber stamp. It's really hard to communicate the importance of self-directed education/learning how to learn when the vast majority of students' formal educational experiences tell them otherwise. Very frustrating but perhaps things are changing ...
School has two competing goals and this will never change:
1. Have the kids learn new things 2. Have the kids reach a desired level of competency
Learning happens where you are at, not where the teacher wants you to be. Every student is at a different place in understanding. It's impossible without 1-on-1 instruction to really maximize learning.
Competency is only determined via testing. Learning doesn't require testing at all, you can just speak to a student to get a good idea if they're making some progress, any progress. Competency? That basically demands a test, because it has a particular bar in mind.
Now students know they need to pass the bar, somehow, but the anxiety of that is going to cause issues with them just trying to learn. This is unfixable though, because the outside pressures demand students have some level of competency otherwise teachers are viewed as failures.
I agree. Imo, #2 is becoming more of an emphasis over time. Teacher don't have much time/energy to pursue #1. Eventually, most of them stop caring and rely on testing metrics because that's what the admins want.
It's amazing what kids can learn if they just spent a little bit of time with a 1-on-1 instructor/advisor. The anxiety you mentioned can be crippling and something I deal with regularly. Even some of the "gifted" kids (perhaps due to the expectations) have trouble avoiding the trap of overindexing on productivity/competency metrics. They're not even self aware of it, just accepts it as normal.
For most kids I have to go through the exercise of separating these two concerns, the learning part and the signaling part, early so they can put things in perspective.
There's also the validity of learning methods, despite what studies may claim, there's no scientific "grand theory of meta-learning", and if ideas are misapplied/misused there's a risk of falling into scientism, which would be just as harmful as economically driven credentialism. At worst it is just the austerity version of education—learn it yourself because we can't afford the school resources to teach/coach/nurture subjects.
There is no grand theory perhaps, but educational research is not totally useless, people have accumulated a lot of insights over the years and some of these "intuitions" are also backed by studies.
Here are some book-length reviews of currently known things about learning/teaching that I found to have a very high signal-to-noise ratio:
https://www.routledge.com/How-Learning-Happens-Seminal-Works...
https://www.routledge.com/How-Teaching-Happens-Seminal-Works...
We’re in dire need of this right now. The number of people that I work with who refuse to pick up new tools and technologies is astounding. If they _do_ try something new, they seem to avoid all but the most basic knowledge of whatever it is, and look at me crosseyed if I suggest going the slightest bit deeper (`git add -p` rather than `git add .`, for example).
I'm sure it varies a lot place to place but I've experienced much more of the opposite in the tech industry. I've heard countless times that we should switch to a different tool because it's newer, from someone who couldn't name a single specific way it's better than the existing tool. I see so much busy work at my employers and in products I use where things get changed just for the sake of change, without getting any better.
I feel like it's a mix of both, depending on how familiar thungs are. It's probably much easier for someone in tech to try moving to a new tool for no real reason than it is for them to learn a bit of CAD and a bit of electronics to make a widget (again, for no real reason).
"Teach to the test" started this a while ago - memorizing and then forgetting bullet points vs engaging more deeply with a subject.
If we're lucky, LLMs force people to put more effort into assignments and grading and then that would help kids learn to learn as well.
I think that's a somewhat unfair framing of the issue, since it isn't a lack of effort from teachers, but rather the pressures from higher levels towards increasing test scores. LLMs don't really change that.
There's also the fact that kids are being taught the very basics, the sorts of things increasingly intelligent models are most likely to be able to solve first. I don't think there's any level of effort that can be put into designing assignments to get around this.
Similar to how teachers haven't really been able to do anything to stop kids from sticking their algebra problems into wolframalpha or other tools besides just making them do the work in class (which then cuts into teaching time).
Beyond a certain point, all that can be done is for teachers to try to instill the importance of practice into students, and for parents to be more proactive in monitoring how their kid is doing their homework.
Maybe we'll see an increase in after-school classes for kids to do homework in, under teacher supervision.
> If we're lucky, LLMs force people to put more effort into assignments and grading and then that would help kids learn to learn as well.
I'm afraid it might be exactly opposite. Having all the knowledge at hand. all the time will lead to knowledge atrophy. Just like it already happens with ability to travel without navigation.
With how rapidly LLMs are improving, I don't know how you would construct assignments that can't be solved relatively quickly by a student feeding it into a bleeding-edge LLM. Especially since teachers often aren't PhDs and are overworked, the idea of every class of students getting handed brand new problems that aren't in the training set feels far-fetched.
I hope somebody figures this out but I don't know what the solution looks like.
Xlation: Classic Corporate America Condescension.
Trust me if google can do something anyone can. They are trying to "define" what "they" "want" from a "compliant workforce"
first off it doesn't seem to be taught at the moment, but also I'm pretty sure that has always been the most important and foundational skill, and it seems like there might be an upper bound for what percentage of people can actually learn it.
Learning is a nontrivial skill even though it has historically been treated as such. It requires an embodied understanding of concepts from basic cognitive psychology, expertise theory, behavioral-affective psychology, metacognition, and more. Until people stop with the platitudes of "learning how to learn is important" and start teaching/training the subject directly as a skill that must be acquired, this will not change.
Simply showing a learner a few slides on spaced retrieval will not cut it.
I'd place this skill, "Learning how to Learn", with Cal Newport's notion of "Deep Work". Part of me wants to say that the latter is a precondition of the former, but I'm not sure that's the case.
https://idriesshahfoundation.org/books/learning-how-to-learn...
Google's top AI scientist says 'learning how to learn' will be next generation's most needed skill
https://techxplore.com/news/2025-09-google-ai-scientist-gene...
One of my favorite pieces on this topic is this talk "Stop Treading Water: Learning to Learn":
https://www.youtube.com/watch?v=j0XmixCsWjs
This is every generation's .ort needed skill
Learning how to learn is the most important skill to acquire in school. If I may add, learning how to learn effectively is a lifetime journey.
Leaning how to think critically is what the next genertion's most needed skill SHOULD be
I think all knowlwdge "work" is bust. Selling knowlwdge will be like selling knitted socks.
People who do stuff will make money
A lot of what people think of as "doing stuff" relies on years or decades of training and experience. When you pay someone to maintain part of your house or to create bespoke furniture or repair your car you're not just paying for labor, you're paying for labor from someone with knowledge and equipment.
Feels like a bit of a useless cliche. Being able to learn has far more to do with iq than any pedagogical technique.
I think there's an element of curiosity that is innate, and that make a huge difference. For those so blessed, learning is fun and a welcome activity (there's an entirely separate issue of actually applying any of knowledge but at least it's a start).
That said, teachers can make all the difference in connecting with students to guide them into discovery vs. cramming facts into their heads to satisfy testing metrics.
Terrible article. Did Hassabis really say ”learning how to learn” is itself a skill, like the article paraphrases it? Surely the skill isn’t that you can learn how to learn, the skill is that you know how to learn. Just like ”learning to ride a bike” isn’t a skill; it’s something you do once, leading to a useful skill.
”Learning how to learn” sounds vaguely insightful just because of the repetition, but if you think for a bit about what it actually means it falls apart.
Imo it's simple, just learn by doing and making stuff.
Care to say more?
As a Curriculum Designer and OG dial-up Millennial I agree wholeheartedly. Too late though. Enlightenment, collaboration, and advancing human experience isn't profitable.
In other words, until one learns how to hammer a nail, it's unreasonable to assume knowledge of how to tell another to do so. AI is no exception. It's speed-running US society's final threads being severed, and okay, sigh, here we go. No, I'm not interested in fixing the problems he's identifying.
My ex had a saying from bench science..."if you're not part of the solution, you're part of the precipitant." That part. Off to go live in a van down by the river...
This generation's most needed skill is being born to rich parents.
But it'll be totally different in the next generation. Trust me, bro, I'm rich - so I must know what I'm talking about.
/s !
Future internet road maps be like:
Join the Generalist bootcamp, it includes big picture of the world and everything, anything you ever need. Full access subscriptions at $1000.
You will Learn the following things:
Analytic philosophy, Mathematical logic, Pure and applied math, Physics, CS, Systems thinking, Engineering(Mech + electronics), creative problem solving And finally one art subject
Beginer Projects: Wafer stage design. Model nano tech projects. Small nuclear fusion reactors. Portable TEM machine.
Pre-req: Just enough maturity. You should be curious, persistence & hardworking. We assume you will practise problem solving till you die.
Outcomes of the bootcamp: Job guaranteed at fortune 500.
Testimonials: We have so many happy customers working for companies having trillion dollar values.
Horror of horrors. Please no.