Wow, a new technology has been introduced that seems to provide nothing but upside. I wonder if there will be any unforeseen consequences that don’t manifest until it’s already too late and an entire generation has had its brains fried. No way to tell, it’s not like that’s ever happened before.
The consequences are already here, from an recent MIT study:
"After four months of AI-assisted writing, participants’ brains showed up to 55 per cent reduced neural connectivity compared to those who wrote independently. More importantly: when AI was taken away, their brains did not recover. The neural engagement patterns did not snap back. The cognitive architecture had been structurally reorganised."https://schooloftheunconformed.substack.com/p/the-sacred-triad-where-and-why-to
Last April my husband and I attended on AI and the Future of Education at the Perimeter Institute for Theoretical Physics and were expecting to hear a lot of hype about how it would revolutionize education. Instead, to our great surprise, the speaker (Derek Muller) emphasized that 1. AI is used to render learning easier, thus bypassing effortful engagement and fundamental "system 1" thinking, 2. the real innovation lies not in using yet another technology, but people, because learning happens in the context of relationships.
Learning = effort + relationship / information
To learn something, we need to use cognitive effort, and we need to do it through human relationships; and since we can only absorb so much information at once, we need to present just enough information to challenge students but not to overwhelm them. Then repeat until mastery, and keep doing it.
Over the last two decades I've repeatedly observed that what propelled my children's learning was not the subject matter per se, or who well the curriclum was designed, but whether they connected with and admired their teachers. My daughter ended up studying French in universtiy because she was deeply inspired by her French teacher; my son pursued medieval studies because of the incredible passion his high school teacher brought to the class.
Passion for learning may ignite through subject matter alone, but it is the human teachers that bring it to life who act as catalysts.
This tracks… just yesterday I was hearing the umpteenth examples of “my kid chose this career because she had a really exciting teacher who taught it in high school”. It’s a very common story… how crazy is it that sometimes we just need someone to teach us something well, to transmit their love of that subject to us?
"Education is not acquiring a stock of ready-made ideas, images, sentiments, beliefs and so forth; it is learning to participate in the conversation of mankind"
Of course, we are all being habituated into a new form of life that is provably bad for all of us. Zooming out from the narrow goal of improving performance on PISA scores; the replacement of teacher time with "more efficient" screen time is just another way of normalizing this form of life.
But I've come to believe that there is a deeper hidden cost, here. We have a chance to signal care to children, through the provision of real human contact (ie the only kind of contact that can facilitate actual caring, see the Care Ethics tradition), and we are sending an entirely different signal altogether, one which might be catastrophic at scale... https://nicholassmyth.substack.com/p/efficiency-and-the-decline-of-care
The folks who run Alpha School are not being totally honest, or rather they’re hyping it a bit. It’s not really “AI.” It’s a textbook on a computer with tests built in. It’s conceptually no different than “IPI,” which was a 60s-70s era individualized math program where kids did workbooks and proceeded at their own pace. IPI was all the rage at one time, but it fell out of favor. (I use that as an example because I learned math from IPI and was arguably a modest success for the program.)
They don’t have “teachers,” but they do have “guides” who explain things when kids get stuck.
And they offer rewards for advancing rapidly — points, prizes and such.
Apparently high IQ kids from high income families who respond to rewards can rip through the elementary school curriculum in two hours a day. Sure, but the same would be true if they just used books.
What you're naming as metabolisation is, I think, the same thing the neuroscience of genuine encounter has been quietly demonstrating for the last two decades — which is that the gap between your daughter's Duolingo proficiency and her proficiency in the wild is not a gap in information but a gap in coupling. Uri Hasson's hyperscanning work shows that genuine human communication produces measurable neural synchrony between persons: meaning does not travel between isolated systems, it emerges in the coupling between them, and that coupling requires physical co-presence in ways that screen-mediated interaction demonstrably fails to replicate. The AI tutor can build the knowledge graph with impressive precision — it can deliver the right content at the right Lexile level, fill the Jenga-tower gaps, run the spaced repetition — but it cannot supply the thing that actually completes learning, which is the moment when a nervous system trained in one context is required to couple with another living nervous system and discovers what it actually knows. Your Italian example is not an edge case or an anomaly; it is the structure of all genuine knowing. The robot can build the vessel. It cannot be the water.
Rabidly anti-AI. It’s too dangerous and humans too flawed to mess around with it at all and think it just a “tool.” That radically underestimates what AI actually is ontologically.
There is a psychological principle that 100 percent abstinence is easier than 98 percent and that applies here.
The dirty little secret of schools is most of the time is spent on classroom management and spurious games, especially at younger ages. Young kids getting through everything in 2 hours isnt really a huge feat, many homeschoolers can do that.
Studies have clearly shown better retention with old fashioned paper and pencil than screens. These kids dont need the whiz bang AI. A few books would have the same result.
All this brings to the forefront is the heritability of IQ and the grossly poor pedagogy of most schools.
For foreign language AI does have a place, as it gives a "good enough" foreign language teacher 24/7.
There's a saying in athletic training that you actually get fitter when you sleep. Training (or any physical exertion) *tears* muscles and breaks them down; you improve because your body rebuilds them a little stronger to resist this. It seems rather obvious that this is true of learning too, just as you need sleep and food to train physically, it's quite possible that the retention part of learning happens elsewhere and that perhaps two hours of intense learning followed by social messing about is a more productive way to go about educating children than seven not-very intense (and dull hours) with less social interaction. OTOH, IIRC, all the kids at that school had bright parents who were probably motivated to learn and these things are genetic and passed on through interactions with those parents, so these aren't typical children, let alone struggling ones. Pretty much any exposure to education would seem to work with them.
I am a certified oldde phartte who uses AI; take everything I say with your preferred measure of salt. In the physical world there are many inert substances, meaning they do not in their natural state react with with other substances to form new ones. Nitrogen and gold are just two among many. In this context they might represent two general levels of innate, but inactive intelligence in humans. Let's take nitrogen as our example (N2). The Haber-Bosch process (TLDR), under very hight temperature and pressure, converts inert and ubundant atmospheric nitrogen to ammonia, which in turn is the basis for fertilizers that make modern intensive agriculture possible. Please do lecture me on the evils of modern agriculture. I am making a point about activating that which is inert. Activating that which is inert requires at least three ingredients, energy, knowledge and a process that works. New technologies give access to new knowledge while disrupting existing processes. AI in education is a powerfully disruptive technology that can and should be mobilized to activate inert minds. Our most urgent problem lies in the fact that AI is proliferating rapidly while too many in the teaching and (dare I say it) the parenting business, need their minds activated first. Who will provide that service for them if they do not do it themselves? It will take energy, absorbing the new knowledge and fitting it to or redesigning the process. I can't hold my breath long enough for it to happen, so my advice is turn AI loose on the kids. Let them make their mistakes, go up creeks without paddles, but as parents and teachers, stay close to band-aid the scrapes and above all, abosrb what the kid-mediated AI can teach us that AI alone cannot.
It's all said with the phrase "industrial education." Do we want our education system to be about producing future workers and consumers, or do we want it to be about becoming who your are? If the latter, Alfa school is the bomb!
I know just as much about Mary Harrington as she does about Alpha school, but I am 100% with her on this one.
And if you want to read about possible new futures where education is about the work of becoming, maybe you're interested in this post:
I agree totally about the 'moving toward' by the learner. I have always noticed this myself in the way the first time you encounter new material it seems large and long and then as you come back to it, your brain having worked on it, it seems to shrink and forms itself into understandable shapes or pieces, at least it does for me. Then, as I also have to teach, there is another level altogether where you can understand something well enough to (hopefully at least) be able to explain it clearly to other people. As for language learning, I think that language specialists (see for example Colin Gorrie's substack e.g.https://www.deadlanguagesociety.com/p/why-people-fail-at-learning-languages ) have views on Duolingo but I think it does help as a practice tool and young children of course are so much more able to pick up new languages. If you moved to Italy, I'm sure your daughter would be fluent within months.
If AI assisted learning can be seen as a robotic one-on-one tutor, as in the Alpha model, I think the focus needs to be on the whole afternoon of freedom that these students have where they can apply the knowledge, or metabolize the knowledge as you mentioned above. That's the way it could be optimal. It's not replacing humans, it's having the human connection and mentoring and metabolizing of knowledge being done after they have ingested the nuts and bolts in the morning. If meaningful afternoon activities aren't in conjunction with their morning learning, it would be less desirable.
This quote from Will Orr-Ewing’s post (Thankyou for sharing it Mary) seems to get to the root of the discussion: What is the goal of tutoring?
“Whereas once the goal was intellectual stimulation, academic excellence for its own sake, or the glorification of God - rarely social refinement! - now the overwhelming majority of tutoring pursuesthin and transactional ends: success in an exam such as the 11+, GCSE or A level.”
I am cautiously supportive of AI as a learning tool within well thought-through ethical boundaries (eg. Research tool = ok; replacement writer = not ok).
I'm even convinced that those who reject it should interact with it first so they understand what they are rejecting and why.
I agree that there is a relational component of learning that can never be replaced by anything else. I think small class sizes are the best way to maximise the relational dimension of learning.
Wow, a new technology has been introduced that seems to provide nothing but upside. I wonder if there will be any unforeseen consequences that don’t manifest until it’s already too late and an entire generation has had its brains fried. No way to tell, it’s not like that’s ever happened before.
The consequences are already here, from an recent MIT study:
"After four months of AI-assisted writing, participants’ brains showed up to 55 per cent reduced neural connectivity compared to those who wrote independently. More importantly: when AI was taken away, their brains did not recover. The neural engagement patterns did not snap back. The cognitive architecture had been structurally reorganised."https://schooloftheunconformed.substack.com/p/the-sacred-triad-where-and-why-to
Last April my husband and I attended on AI and the Future of Education at the Perimeter Institute for Theoretical Physics and were expecting to hear a lot of hype about how it would revolutionize education. Instead, to our great surprise, the speaker (Derek Muller) emphasized that 1. AI is used to render learning easier, thus bypassing effortful engagement and fundamental "system 1" thinking, 2. the real innovation lies not in using yet another technology, but people, because learning happens in the context of relationships.
Learning = effort + relationship / information
To learn something, we need to use cognitive effort, and we need to do it through human relationships; and since we can only absorb so much information at once, we need to present just enough information to challenge students but not to overwhelm them. Then repeat until mastery, and keep doing it.
You can read our essay reporting on this talk here: Learning Fast, and Slow: Why AI will not revolutionize education https://schooloftheunconformed.substack.com/p/learning-fast-and-slow-why-ai-will
Over the last two decades I've repeatedly observed that what propelled my children's learning was not the subject matter per se, or who well the curriclum was designed, but whether they connected with and admired their teachers. My daughter ended up studying French in universtiy because she was deeply inspired by her French teacher; my son pursued medieval studies because of the incredible passion his high school teacher brought to the class.
Passion for learning may ignite through subject matter alone, but it is the human teachers that bring it to life who act as catalysts.
This tracks… just yesterday I was hearing the umpteenth examples of “my kid chose this career because she had a really exciting teacher who taught it in high school”. It’s a very common story… how crazy is it that sometimes we just need someone to teach us something well, to transmit their love of that subject to us?
Michael Oakeshott:
"Education is not acquiring a stock of ready-made ideas, images, sentiments, beliefs and so forth; it is learning to participate in the conversation of mankind"
He's right, of course.
Of course, we are all being habituated into a new form of life that is provably bad for all of us. Zooming out from the narrow goal of improving performance on PISA scores; the replacement of teacher time with "more efficient" screen time is just another way of normalizing this form of life.
But I've come to believe that there is a deeper hidden cost, here. We have a chance to signal care to children, through the provision of real human contact (ie the only kind of contact that can facilitate actual caring, see the Care Ethics tradition), and we are sending an entirely different signal altogether, one which might be catastrophic at scale... https://nicholassmyth.substack.com/p/efficiency-and-the-decline-of-care
The folks who run Alpha School are not being totally honest, or rather they’re hyping it a bit. It’s not really “AI.” It’s a textbook on a computer with tests built in. It’s conceptually no different than “IPI,” which was a 60s-70s era individualized math program where kids did workbooks and proceeded at their own pace. IPI was all the rage at one time, but it fell out of favor. (I use that as an example because I learned math from IPI and was arguably a modest success for the program.)
They don’t have “teachers,” but they do have “guides” who explain things when kids get stuck.
And they offer rewards for advancing rapidly — points, prizes and such.
Apparently high IQ kids from high income families who respond to rewards can rip through the elementary school curriculum in two hours a day. Sure, but the same would be true if they just used books.
What you're naming as metabolisation is, I think, the same thing the neuroscience of genuine encounter has been quietly demonstrating for the last two decades — which is that the gap between your daughter's Duolingo proficiency and her proficiency in the wild is not a gap in information but a gap in coupling. Uri Hasson's hyperscanning work shows that genuine human communication produces measurable neural synchrony between persons: meaning does not travel between isolated systems, it emerges in the coupling between them, and that coupling requires physical co-presence in ways that screen-mediated interaction demonstrably fails to replicate. The AI tutor can build the knowledge graph with impressive precision — it can deliver the right content at the right Lexile level, fill the Jenga-tower gaps, run the spaced repetition — but it cannot supply the thing that actually completes learning, which is the moment when a nervous system trained in one context is required to couple with another living nervous system and discovers what it actually knows. Your Italian example is not an edge case or an anomaly; it is the structure of all genuine knowing. The robot can build the vessel. It cannot be the water.
Rabidly anti-AI. It’s too dangerous and humans too flawed to mess around with it at all and think it just a “tool.” That radically underestimates what AI actually is ontologically.
There is a psychological principle that 100 percent abstinence is easier than 98 percent and that applies here.
AI learning is the contemporary evolution of generalized computer learning of the last 40 years that has failed to help our students.
The main beneficiaries during that time have been computer hardware and software companies, not students or teachers.
The dirty little secret of schools is most of the time is spent on classroom management and spurious games, especially at younger ages. Young kids getting through everything in 2 hours isnt really a huge feat, many homeschoolers can do that.
Studies have clearly shown better retention with old fashioned paper and pencil than screens. These kids dont need the whiz bang AI. A few books would have the same result.
All this brings to the forefront is the heritability of IQ and the grossly poor pedagogy of most schools.
For foreign language AI does have a place, as it gives a "good enough" foreign language teacher 24/7.
There's a saying in athletic training that you actually get fitter when you sleep. Training (or any physical exertion) *tears* muscles and breaks them down; you improve because your body rebuilds them a little stronger to resist this. It seems rather obvious that this is true of learning too, just as you need sleep and food to train physically, it's quite possible that the retention part of learning happens elsewhere and that perhaps two hours of intense learning followed by social messing about is a more productive way to go about educating children than seven not-very intense (and dull hours) with less social interaction. OTOH, IIRC, all the kids at that school had bright parents who were probably motivated to learn and these things are genetic and passed on through interactions with those parents, so these aren't typical children, let alone struggling ones. Pretty much any exposure to education would seem to work with them.
I am a certified oldde phartte who uses AI; take everything I say with your preferred measure of salt. In the physical world there are many inert substances, meaning they do not in their natural state react with with other substances to form new ones. Nitrogen and gold are just two among many. In this context they might represent two general levels of innate, but inactive intelligence in humans. Let's take nitrogen as our example (N2). The Haber-Bosch process (TLDR), under very hight temperature and pressure, converts inert and ubundant atmospheric nitrogen to ammonia, which in turn is the basis for fertilizers that make modern intensive agriculture possible. Please do lecture me on the evils of modern agriculture. I am making a point about activating that which is inert. Activating that which is inert requires at least three ingredients, energy, knowledge and a process that works. New technologies give access to new knowledge while disrupting existing processes. AI in education is a powerfully disruptive technology that can and should be mobilized to activate inert minds. Our most urgent problem lies in the fact that AI is proliferating rapidly while too many in the teaching and (dare I say it) the parenting business, need their minds activated first. Who will provide that service for them if they do not do it themselves? It will take energy, absorbing the new knowledge and fitting it to or redesigning the process. I can't hold my breath long enough for it to happen, so my advice is turn AI loose on the kids. Let them make their mistakes, go up creeks without paddles, but as parents and teachers, stay close to band-aid the scrapes and above all, abosrb what the kid-mediated AI can teach us that AI alone cannot.
It's all said with the phrase "industrial education." Do we want our education system to be about producing future workers and consumers, or do we want it to be about becoming who your are? If the latter, Alfa school is the bomb!
I know just as much about Mary Harrington as she does about Alpha school, but I am 100% with her on this one.
And if you want to read about possible new futures where education is about the work of becoming, maybe you're interested in this post:
https://writerbytechnicality.substack.com/p/stories-of-a-world-to-come?r=3anz55
I agree totally about the 'moving toward' by the learner. I have always noticed this myself in the way the first time you encounter new material it seems large and long and then as you come back to it, your brain having worked on it, it seems to shrink and forms itself into understandable shapes or pieces, at least it does for me. Then, as I also have to teach, there is another level altogether where you can understand something well enough to (hopefully at least) be able to explain it clearly to other people. As for language learning, I think that language specialists (see for example Colin Gorrie's substack e.g.https://www.deadlanguagesociety.com/p/why-people-fail-at-learning-languages ) have views on Duolingo but I think it does help as a practice tool and young children of course are so much more able to pick up new languages. If you moved to Italy, I'm sure your daughter would be fluent within months.
If AI assisted learning can be seen as a robotic one-on-one tutor, as in the Alpha model, I think the focus needs to be on the whole afternoon of freedom that these students have where they can apply the knowledge, or metabolize the knowledge as you mentioned above. That's the way it could be optimal. It's not replacing humans, it's having the human connection and mentoring and metabolizing of knowledge being done after they have ingested the nuts and bolts in the morning. If meaningful afternoon activities aren't in conjunction with their morning learning, it would be less desirable.
This quote from Will Orr-Ewing’s post (Thankyou for sharing it Mary) seems to get to the root of the discussion: What is the goal of tutoring?
“Whereas once the goal was intellectual stimulation, academic excellence for its own sake, or the glorification of God - rarely social refinement! - now the overwhelming majority of tutoring pursuesthin and transactional ends: success in an exam such as the 11+, GCSE or A level.”
I am cautiously supportive of AI as a learning tool within well thought-through ethical boundaries (eg. Research tool = ok; replacement writer = not ok).
I'm even convinced that those who reject it should interact with it first so they understand what they are rejecting and why.
I agree that there is a relational component of learning that can never be replaced by anything else. I think small class sizes are the best way to maximise the relational dimension of learning.