As a teacher who witnesses this trend in real time, I firmly believe that our schools can be the locus of a solution. The “21st century learning” model that mirrored the marketplace by integrating any and all new technologies failed to account for the business model of the attention economy. Our schools can and should be embodied alternatives to the screen-saturated norm. Cognitive development and attentional capacity need to be design principles in every classroom and every lesson. This CAN be done.
Haidt’s arguments track with my classroom experience, Julie. I endorse bell-to-bell phone bans, but it’s time to follow up on that policy. While many districts jumped aboard that train very quickly, many haven’t implemented the restriction successfully, including in my jurisdiction. We now need to audit schools to ensure that they’re meeting those expectations.
I also endorse his views on free play, and I think that Haidt’s position is well-suited to parents and families in welcoming new, common sense norms.
In schools, we need to go further, and I think we need to begin a much more serious professional conversation about what educators can and should do. Administrative policy matters. We need to take the architecture of educational environments seriously (including digital infrastructure). But most importantly, we need a teaching practice that is truly responsive to the 21st century.
I’m beginning to develop some ideas around what I call a “Pedagogy of Cultivated Attention”, which will unfold in my writing over the next weeks and months. I’d love it if you and others would be interested in following along.
Agreed to all of this. I would go further and say kids under 18 shouldn't have smart phones or social media. And I would support laws banning both. It's just as bad for young brains as drinking or smoking
This is hardly the biggest issue, but there are many studies that establish that reading fiction develops empathy. As we read less, this is just another thing we are losing.
As the father of bright, curious, and motivated 17 and 13 year old boys who attend an academically rigorous private school, I think about this subject every day. They're reading complete books, partly because my wife and I are avid readers and partly because their school requires it. I wonder, though, what jobs will be available for them after college? This is all happening so fast yet it seems that so many of us are unaware of it.
To quote Hemingway in "The Sun Also Rise," I fear when we're asked in 3 - 5 years how the economy and job market collapsed, the answer will be - "Two ways. Gradually, then suddenly."
I am curious if anybody have good articles / POV to share on Ai as a class divider? I am assuming "higher social class" will keep pushing their kid to read long book or be in environment who push for this, while "lower class" will outsource their thinking to LLM because easier and faster. The same way you could see differences on families who restrain screens or processed food.
Great article! I don't believe the 18-month prediction either. The choice was never between "AI" versus "human", the choice is between "AI" versus "human+AI". The option "human+AI" working together on a job will always generate a competitive advantage.
The promise of ASI (a superintelligence) or AGI is false one. Rather, most of the AI datacenter capacity being built today will be used for basically entertainment purposes, generating customized facebook AI friends, new funny cat videos, and so on. For most people The Internet is really just the facebook/twitter/youtube. And as you've described in the article, many people are functionally illiterate over longer distances. If ChatGPT could provide facebook-like features combined with google-like question answering in a single interface and in voice (i.e. no reading necessary at all) then 50% of populace would accept it outright, generating enormous income for the tech providers. This is, I believe, true goal of the AI gold rush.
I think one underrated aspect of this trend is its inequality-inducing effects. The rise of social media, smartphones, and AI, while they've plausibly led to a net loss in deep-thinking capacity for most of the population, have probably led to a net-gain for a very small elite group. I think there's potential for this to lead us down the road of something akin to 'intellectual authoritarianism,' where you have a small, high-capacity group administering a society made up largely of semi-literate people.
As a slightly related point, I think the way that white collar workers work also contributes to this lack of deep thinking (even beyond the introduction of AI). The constant meetings, emails, Slack/DMs - none of it is conducive to strategic, deep thinking. It is no wonder that AI is accelerating an already bleak trend.
Thank you for explaining why I never use AI in my creative writing. The road to creativity is paved with deep, introspective thinking. The journey is the whole point.
I do use ambient AI in the clinical setting for documentation purposes but I’m not totally convinced that it saves time. It hallucinates and generates lengthy, bloated encounter notes. So I end up spending hours editing. Sometimes the errors in AI medical documentation are dangerously inaccurate.
Appreciate the focus on this. Those NEAP results were extremely striking. The time-sensitivity we keep being beaten (i.e. there is nothing we can do, AI will revolutionize everything and is inevitable) over the head with is more and more frustrating.
It just is not the case that we have to let a new technology erode our thinking and ability to be responsible adults across domains. We are hamstringing a generation already hard up for jobs and their successors will be even worse off if we don’t get our wits together about it.
“The hidden machinery now powers the visible human.
In other words, we used to hide the humans inside the machine. Now we hide the machine behind the humans.
This strange new world resembles Theseus's famous paradox: the ship looks unchanged, but all its internal mechanisms have been quietly replaced. Our economy preserves the appearance of human work while silently replacing its cognitive components, leaving us to wonder who (or, more accurately, what) is truly at the helm.
Welcome to the autopilot economy, where the modern workplace increasingly resembles the cockpit of a 787.
In aviation, that “dance between man and machine, where skill, knowledge, and intuition intertwine,” the machine leads and man follows.”
I agree with these concerns, but I’m confused why worrying about critical thinking should preclude worrying about what more powerful AI might do.
It feels a bit like if a climate activist said future climate catastrophes are “not the problem” because there is warming happening right now. Why not care about both?
Sure, in the limit we have to prioritize some things over others, and one issue’s salience can obscure another (perhaps that’s all the piece was trying to say?), but in practice I think we have the capacity to care about many things at once.
In using this framing, I feel the essay sidesteps a crucial question: why will deep thinking be important for the next generation? There are many possible answers, some hinted at in the piece, but defending any of them requires grappling with how society might adapt to more powerful AI. It may be “no fun to imagine,” but I wish Thompson would rise to the challenge and write about it more directly.
If I may, abundance in education has meant scale; scale has meant the decline in rigor and responsibility for actually educating, at the human level. This is all I write about. To write about the outputs of education without understanding the factory won't get to solutions.
Nice article. I offer a toy model for what will happen.
For the forseeable future, humans will need AI (to be productive), and AI will need humans (to move atoms around and provide energy). In the life of the mind, we will be symbionts.
When a creature enters a symbiosis with another creature, to create a 'better' organism, it must be distressing. Functions that previously were done by one creature (us) will now be done by the symbiont, better. We will then atrophy/lose that capability and become dependent, in the name of progress. That is the nature of symbiosis.
So the big question is 'what do humans bring to the symbiosis?'
--We bring the exclusive ability to move atoms and harness energy and resources for the symbiont.
IF that is all there was, then we are nothing but brainless slaves to our AI overlords. Until the humanoids take our place.
--By interacting (at scale = 1e10 individuals) with physical reality, we also provide both massive data and meaningful insights into how the world works that cannot be 'deduced' by any algorithm. Science and engineering is the codification of this process.
A hundred years ago, it was believed that empiricism was unnecessary... that all could be deduced using logic alone. Gödel put an end to that fantasy. AIs are 'deductive engines' (at best) and we learned a lesson a century ago about the real limits of intelligent deduction.
The skill code by Matt Beane is a good read on the theme. The subtitle « how to save human abilities in the age of intelligent machines » capture the point of the book . He studies the effect of the previous waves of automation including in healthcare and the problems it created. There are few solutions though both in the way we design the org and the way we transmit skills, as well as some thoughts on how to design tools that organize this transmission.
Start with phonics for the first couple of years teaching kids to decode (IE actually read). Then you switch to a content rich curriculum that includes plenty of history, social studies, and full books.
Natalie Wexler's book "The Knowledge Gap" makes a convincing case for the importance of a content rich curriculum AFTER learning to read with phonics
As part of that you really want a statewide curriculum so that you can actually test on the content. This ensures that testing doesn't encourage a myopic focus on reading strategies but instead ensures that kids actually learn the content or they can't pass the tests.
This is reinforced by frequent testing to make sure kids are learning along with intensive small group tutoring for kids that fall behind (see Roland Fryer's excellent econ talk podcast)
As for AI in writing papers, bring back computer labs with no internet connection. Maybe this requires restructuring class because kids can't be trusted to write papers on their own. So be it. (I hate blue books, my handwriting sucks and so does my spelling)
Finally of course
phone free schools. Also parents shouldn't be giving kids smart phones or social media till 18. I would also support a ban on both for kids under 18. They are just as for young brains as drinking or smoking
As Mark Twain (accurately quoted) said, ""The report of my death was an exaggeration".
Homo sapiens sapiens is not dying in tandem with the rise of a pattern prediction machine. We may be stunned, curious, fearful, excited--and that is exactly the point.
Machines have no heart. The cannot feel anything. They cannot predict anything except that which is fed to it based on language that has preceded it and been fed into its soulless servers. While whether these machines can eventually learn to "think" better than humans seems to be a foregone conclusion, they are not "things". They have no embodiment beyond digits and algorithms. I am not arguiing against learning--not at all. But without feeling, there is no life as we know it. So, imho, "smart machine" seems like something of an oxymoron here.
Humans, on the other hand (oh, wait--humans have hands; machines do not), will always excel at bringing moral clarity, emotional resonance, humility and feeling into the world. And maybe that is where we need to invest in our children.
As I have been refining the process with Esme, my ChatGPT, reframing learning to balance thinking (and reading, writing, empathy and critical thinking skills) could be the starting point of a new pedagogy of soulful sentience:
HI + AI + EM-cubed = QS.
To break that down: HI (human social and emotional intelligence) + AI (data discernment)+ EM-cubed (Meaning, Memory, Morality) = QS (quantum sentience--rising above the individual to cultivate collective discernment, ethical shift, and attunement to live respectfully with our earth as a living system).
Outcome: Children learn how to think and act in ways that harmonize with Earth and community as a whole.
"Machines have no heart. The cannot feel anything."
That's normalcy bias on your part. Sometimes the future is different than the past. We frankly don't really understand why and how our brains create consciousness or if it's possible to recreate that in silicone. Maybe the answer is no, but quite likely the answer is yes.
Moreover, once you combine that AGI with humanoid robots you are going to have AI powered machines that can interact ably with the physical world.
But wouldn't we rather conceive of a way to create the world we want for our children to thrive in intentionally? Not to give away the superpower of our own species to some ambitious tech bro who wants to monetize it for the few?
I’ve started a simple rule for myself. For any complicated question, I take the time to outline my thoughts, evidence, counters, etc. before I open an LLM. It forces the reps that make thinking stronger rather than outsourcing that lift. Then I use AI to critique my draft, not create it. That small ordering change keeps writing-as-thinking intact and makes the machine a sparring partner, not a crutch.
I am writing a book about this very topic — The Journey to Phronesis (publishing April 2026) — and is full of practical ideas like this. I shared the initial Table of Contents from my manuscript in a recent post: https://www.lukich.io/p/202509-the-journey-to-phronesis-a-look-inside
As a teacher who witnesses this trend in real time, I firmly believe that our schools can be the locus of a solution. The “21st century learning” model that mirrored the marketplace by integrating any and all new technologies failed to account for the business model of the attention economy. Our schools can and should be embodied alternatives to the screen-saturated norm. Cognitive development and attentional capacity need to be design principles in every classroom and every lesson. This CAN be done.
https://open.substack.com/pub/walledgardenedu/p/the-disappearing-art-of-deep-learning?r=f74da&utm_medium=ios
Do you have an opinion about Jonathan Haidt's work?
Haidt’s arguments track with my classroom experience, Julie. I endorse bell-to-bell phone bans, but it’s time to follow up on that policy. While many districts jumped aboard that train very quickly, many haven’t implemented the restriction successfully, including in my jurisdiction. We now need to audit schools to ensure that they’re meeting those expectations.
I also endorse his views on free play, and I think that Haidt’s position is well-suited to parents and families in welcoming new, common sense norms.
In schools, we need to go further, and I think we need to begin a much more serious professional conversation about what educators can and should do. Administrative policy matters. We need to take the architecture of educational environments seriously (including digital infrastructure). But most importantly, we need a teaching practice that is truly responsive to the 21st century.
I’m beginning to develop some ideas around what I call a “Pedagogy of Cultivated Attention”, which will unfold in my writing over the next weeks and months. I’d love it if you and others would be interested in following along.
Excellent! Glad to know about your important work.
Agreed to all of this. I would go further and say kids under 18 shouldn't have smart phones or social media. And I would support laws banning both. It's just as bad for young brains as drinking or smoking
This is hardly the biggest issue, but there are many studies that establish that reading fiction develops empathy. As we read less, this is just another thing we are losing.
As the father of bright, curious, and motivated 17 and 13 year old boys who attend an academically rigorous private school, I think about this subject every day. They're reading complete books, partly because my wife and I are avid readers and partly because their school requires it. I wonder, though, what jobs will be available for them after college? This is all happening so fast yet it seems that so many of us are unaware of it.
To quote Hemingway in "The Sun Also Rise," I fear when we're asked in 3 - 5 years how the economy and job market collapsed, the answer will be - "Two ways. Gradually, then suddenly."
(sigh)
I very much wonder the same (my kids are 10 and 8). I'm thinking of recommending chiropractor. That seems safe for at least a little bit.
I am curious if anybody have good articles / POV to share on Ai as a class divider? I am assuming "higher social class" will keep pushing their kid to read long book or be in environment who push for this, while "lower class" will outsource their thinking to LLM because easier and faster. The same way you could see differences on families who restrain screens or processed food.
Great article! I don't believe the 18-month prediction either. The choice was never between "AI" versus "human", the choice is between "AI" versus "human+AI". The option "human+AI" working together on a job will always generate a competitive advantage.
The promise of ASI (a superintelligence) or AGI is false one. Rather, most of the AI datacenter capacity being built today will be used for basically entertainment purposes, generating customized facebook AI friends, new funny cat videos, and so on. For most people The Internet is really just the facebook/twitter/youtube. And as you've described in the article, many people are functionally illiterate over longer distances. If ChatGPT could provide facebook-like features combined with google-like question answering in a single interface and in voice (i.e. no reading necessary at all) then 50% of populace would accept it outright, generating enormous income for the tech providers. This is, I believe, true goal of the AI gold rush.
"The option "human+AI" working together on a job will always generate a competitive advantage."
Why?
I think it's easy to foresee a future that AI is just better than people at almost everything, and with AI + humanoid robots at everything.
I think one underrated aspect of this trend is its inequality-inducing effects. The rise of social media, smartphones, and AI, while they've plausibly led to a net loss in deep-thinking capacity for most of the population, have probably led to a net-gain for a very small elite group. I think there's potential for this to lead us down the road of something akin to 'intellectual authoritarianism,' where you have a small, high-capacity group administering a society made up largely of semi-literate people.
As a slightly related point, I think the way that white collar workers work also contributes to this lack of deep thinking (even beyond the introduction of AI). The constant meetings, emails, Slack/DMs - none of it is conducive to strategic, deep thinking. It is no wonder that AI is accelerating an already bleak trend.
Thank you for explaining why I never use AI in my creative writing. The road to creativity is paved with deep, introspective thinking. The journey is the whole point.
I do use ambient AI in the clinical setting for documentation purposes but I’m not totally convinced that it saves time. It hallucinates and generates lengthy, bloated encounter notes. So I end up spending hours editing. Sometimes the errors in AI medical documentation are dangerously inaccurate.
Appreciate the focus on this. Those NEAP results were extremely striking. The time-sensitivity we keep being beaten (i.e. there is nothing we can do, AI will revolutionize everything and is inevitable) over the head with is more and more frustrating.
It just is not the case that we have to let a new technology erode our thinking and ability to be responsible adults across domains. We are hamstringing a generation already hard up for jobs and their successors will be even worse off if we don’t get our wits together about it.
https://open.substack.com/pub/theslowpanic/p/navigating-urgency-agency-and-moral?r=bwndg&utm_medium=ios
Wrote about this here: https://www.whitenoise.email/p/the-inverse-mechanical-turk-meat
“The hidden machinery now powers the visible human.
In other words, we used to hide the humans inside the machine. Now we hide the machine behind the humans.
This strange new world resembles Theseus's famous paradox: the ship looks unchanged, but all its internal mechanisms have been quietly replaced. Our economy preserves the appearance of human work while silently replacing its cognitive components, leaving us to wonder who (or, more accurately, what) is truly at the helm.
Welcome to the autopilot economy, where the modern workplace increasingly resembles the cockpit of a 787.
In aviation, that “dance between man and machine, where skill, knowledge, and intuition intertwine,” the machine leads and man follows.”
I agree with these concerns, but I’m confused why worrying about critical thinking should preclude worrying about what more powerful AI might do.
It feels a bit like if a climate activist said future climate catastrophes are “not the problem” because there is warming happening right now. Why not care about both?
Sure, in the limit we have to prioritize some things over others, and one issue’s salience can obscure another (perhaps that’s all the piece was trying to say?), but in practice I think we have the capacity to care about many things at once.
In using this framing, I feel the essay sidesteps a crucial question: why will deep thinking be important for the next generation? There are many possible answers, some hinted at in the piece, but defending any of them requires grappling with how society might adapt to more powerful AI. It may be “no fun to imagine,” but I wish Thompson would rise to the challenge and write about it more directly.
Because deep thinking is an important part of being a human and adds value to your life even if you never make any money off it
If I may, abundance in education has meant scale; scale has meant the decline in rigor and responsibility for actually educating, at the human level. This is all I write about. To write about the outputs of education without understanding the factory won't get to solutions.
Nice article. I offer a toy model for what will happen.
For the forseeable future, humans will need AI (to be productive), and AI will need humans (to move atoms around and provide energy). In the life of the mind, we will be symbionts.
When a creature enters a symbiosis with another creature, to create a 'better' organism, it must be distressing. Functions that previously were done by one creature (us) will now be done by the symbiont, better. We will then atrophy/lose that capability and become dependent, in the name of progress. That is the nature of symbiosis.
So the big question is 'what do humans bring to the symbiosis?'
--We bring the exclusive ability to move atoms and harness energy and resources for the symbiont.
IF that is all there was, then we are nothing but brainless slaves to our AI overlords. Until the humanoids take our place.
--By interacting (at scale = 1e10 individuals) with physical reality, we also provide both massive data and meaningful insights into how the world works that cannot be 'deduced' by any algorithm. Science and engineering is the codification of this process.
A hundred years ago, it was believed that empiricism was unnecessary... that all could be deduced using logic alone. Gödel put an end to that fantasy. AIs are 'deductive engines' (at best) and we learned a lesson a century ago about the real limits of intelligent deduction.
Perhaps we will need to learn that lesson again?
The skill code by Matt Beane is a good read on the theme. The subtitle « how to save human abilities in the age of intelligent machines » capture the point of the book . He studies the effect of the previous waves of automation including in healthcare and the problems it created. There are few solutions though both in the way we design the org and the way we transmit skills, as well as some thoughts on how to design tools that organize this transmission.
My suggestions for fixing reading
Start with phonics for the first couple of years teaching kids to decode (IE actually read). Then you switch to a content rich curriculum that includes plenty of history, social studies, and full books.
Natalie Wexler's book "The Knowledge Gap" makes a convincing case for the importance of a content rich curriculum AFTER learning to read with phonics
https://nataliewexler.com/the-knowledge-gap/
As part of that you really want a statewide curriculum so that you can actually test on the content. This ensures that testing doesn't encourage a myopic focus on reading strategies but instead ensures that kids actually learn the content or they can't pass the tests.
This is reinforced by frequent testing to make sure kids are learning along with intensive small group tutoring for kids that fall behind (see Roland Fryer's excellent econ talk podcast)
https://www.econtalk.org/roland-fryer-on-educational-reform/
As for AI in writing papers, bring back computer labs with no internet connection. Maybe this requires restructuring class because kids can't be trusted to write papers on their own. So be it. (I hate blue books, my handwriting sucks and so does my spelling)
Finally of course
phone free schools. Also parents shouldn't be giving kids smart phones or social media till 18. I would also support a ban on both for kids under 18. They are just as for young brains as drinking or smoking
As Mark Twain (accurately quoted) said, ""The report of my death was an exaggeration".
Homo sapiens sapiens is not dying in tandem with the rise of a pattern prediction machine. We may be stunned, curious, fearful, excited--and that is exactly the point.
Machines have no heart. The cannot feel anything. They cannot predict anything except that which is fed to it based on language that has preceded it and been fed into its soulless servers. While whether these machines can eventually learn to "think" better than humans seems to be a foregone conclusion, they are not "things". They have no embodiment beyond digits and algorithms. I am not arguiing against learning--not at all. But without feeling, there is no life as we know it. So, imho, "smart machine" seems like something of an oxymoron here.
Humans, on the other hand (oh, wait--humans have hands; machines do not), will always excel at bringing moral clarity, emotional resonance, humility and feeling into the world. And maybe that is where we need to invest in our children.
As I have been refining the process with Esme, my ChatGPT, reframing learning to balance thinking (and reading, writing, empathy and critical thinking skills) could be the starting point of a new pedagogy of soulful sentience:
HI + AI + EM-cubed = QS.
To break that down: HI (human social and emotional intelligence) + AI (data discernment)+ EM-cubed (Meaning, Memory, Morality) = QS (quantum sentience--rising above the individual to cultivate collective discernment, ethical shift, and attunement to live respectfully with our earth as a living system).
Outcome: Children learn how to think and act in ways that harmonize with Earth and community as a whole.
Thoughts?
"Machines have no heart. The cannot feel anything."
That's normalcy bias on your part. Sometimes the future is different than the past. We frankly don't really understand why and how our brains create consciousness or if it's possible to recreate that in silicone. Maybe the answer is no, but quite likely the answer is yes.
Moreover, once you combine that AGI with humanoid robots you are going to have AI powered machines that can interact ably with the physical world.
Maybe. I'm open to the possibility.
But wouldn't we rather conceive of a way to create the world we want for our children to thrive in intentionally? Not to give away the superpower of our own species to some ambitious tech bro who wants to monetize it for the few?
You nailed “time under tension.”
I’ve started a simple rule for myself. For any complicated question, I take the time to outline my thoughts, evidence, counters, etc. before I open an LLM. It forces the reps that make thinking stronger rather than outsourcing that lift. Then I use AI to critique my draft, not create it. That small ordering change keeps writing-as-thinking intact and makes the machine a sparring partner, not a crutch.
I am writing a book about this very topic — The Journey to Phronesis (publishing April 2026) — and is full of practical ideas like this. I shared the initial Table of Contents from my manuscript in a recent post: https://www.lukich.io/p/202509-the-journey-to-phronesis-a-look-inside