So let’s get one thing out of the way: I think “AI literacy” is a dangerous device of neoliberal education and it deserves to be dismissed out of hand.
Ugh yes. Baudelaire called photography “art’s most mortal enemy.” Cameras stole the meaningfulness of skin-pore level realism and painters started leaning into vibes and motion, impressionism etc. Walter Benjamin said photos sucked the “aura” out of representation. I suppose others will say it forced artists to explore new frontiers. I wonder if AI will do the same for writing.
Substack is kinda the problem too (even tho it’s how I make a living). I spend less time interacting with ppl directly cuz I’m writing 1,500 words to strangers and checking the open rate.
Kids can’t read facial expressions because they’re spending too much time on screens. A study just showed that if you get them off screens for five days, they get better at understanding emotions. That’s insane.
AI seems to us like a new special evil, but it won’t to the generation that grows up with it. We’re gonna look like old ppl yelling at the sky. I mean if used carefully it has the potential to make people more literate. Like if you use at as advanced google, ask it to challenge your writing etc. I’ve done that out of curiosity and got interesting results that made me think or learned something new. But of course most ppl want it to do the work for them, like you say. I dunno. We live in interesting times. And a year from now the conversation will be different, probably more intense.
Oh, I accepted my identity as old-lady-sky-yeller a long time ago :)
To be really honest, I am still expecting the bubble to burst. The hype has been so intense that it will take a while for those people to be able to admit that the rapture hadn't come, but the venture capital can't last, and the commercial models are getting worse, and so are the court cases.
I'm not sure the major companies have much time left -- with the exception of Microsoft Copilot and Google Gemini, but everyone hates their LLMs and they are only going to be ok because they do other things that actually do turn a profit.
Someone I know at Google said that, when looking for new hires, he looks for people **not** using AI in their communications. Similarly, I’m pretty sure the VCs hyping these products — telling us we should bet our future on them — would never themselves rely on AI to decide whom to fund because, you know, that’s a more nuanced decision. So far the only uses of AI that seem to impress people is the masking of weak literacy or English as a second language — which, ironically, only ends up drawing attention to the masking and conflating a weak education with ESL, and both as some kind of intellectual “deficiency”— and relationship advice, whose very “revelatory wisdom” says more about the user’s lack of social support systems than about the actual insight that can be culled from aggregating a bunch of human behavioral and emotional patterns into pop psychology short hands. To me this is just environmental extractivism turned inward, onto ourselves. Mental fracking, if you will. I hope you won’t.
Great think piece! Glad I’m not starting out in any school just now. Already have enough challenge w typing my thoughts on a tiny screen and hoping some of my besties will enjoy and/or reply to my rambles!
At this point, i am so old that I’ve seen this story play out more times than I can count. First, it begins with excitement whispers in the corners of conferences, white papers, a few bold demos. Then come the skeptics, loud and certain, armed with their usual dismissals: “This is just a fad.” “It’ll never scale.” “It’s a toy, not a tool.” And then… the world quietly reorganizes around the thing they scoffed at.
It happened with the Internet. In the early ‘90s, it was derided as a curiosity for academics and message board fanatics. It was never going to be “a serious platform for business.” Fast forward, and the internet now underpins the global economy, powers entire governments, and mediates everything from love to war.
Then came online shopping. Oh, the eye-rolls. “Who would buy clothes or groceries online?” they said. “People want to see and touch what they buy.” Today, e-commerce is not just convenient it’s essential infrastructure. Amazon is a utility. The mall is an artifact.
Streaming media? Too glitchy. Low quality. Not a threat to broadcast. And yet, it turned TV networks upside-down, killed cable for a generation, and gave rise to a golden age of on-demand storytelling.
Social media? A waste of time for teenagers. Not serious. Not sustainable. Now it drives elections, brands, markets, and cultural movements. It is quite literally the digital pulse of the world.
Cloud computing was laughed off as insecure and inefficient. “Why would anyone give up control of their own servers?” they asked. Now, not only is the cloud secure, it’s more resilient, more scalable, and more cost-effective than most local infrastructure. It’s the foundation of modern IT.
Open source was once the fringe the realm of hobbyists, anarchists, and “not real” developers. And yet, Linux, Python, Kubernetes, and hundreds of open-source tools now run the majority of our world’s digital infrastructure.
And remember smartphones? The original iPhone was called a luxury. No physical keyboard? Limited battery? Too expensive? Yet within five years, the smartphone had become the most important personal device in history an always-on extension of our minds and lives.
Cryptocurrencies dismissed repeatedly as bubbles, scams, or toys for libertarians have shown remarkable persistence. Volatile? Yes. Complex? Definitely. But here they are, still standing, still evolving, still reshaping finance, asset ownership, and global policy discussions.
Digital compositing and CGI? Too fake. Not “true cinema.” And yet, they’ve become essential tools in storytelling, making the impossible tangible and the fantastic real. Try making a blockbuster without them.
Robotics? Slow. Over-hyped. “We’re not in the Jetsons.” Yet they now quietly run warehouses, perform surgery, inspect pipelines, explore other planets, and assemble your car with terrifying precision.
AI and machine learning? Every decade has its skeptics those who mock each wave as a mirage. But today, AI filters spam, drives cars, predicts pandemics, composes music, writes prose, and diagnoses diseases. Dismissing it now is like dismissing electricity in the age of the telegraph.
Blogging and self-publishing were mocked as amateur-hour. But they democratized voice and gave rise to entire new economies of thought, from indie journalism to niche expertise.
Podcasts were called “radio for nerds.” And yet, they’ve become a dominant storytelling medium, rich with intimacy and nuance, commanding billions of hours of attention.
Containerization was seen as a niche obsession unnecessarily complicated. Today, Docker and Kubernetes are industry standards, transforming how applications are deployed and scaled.
Even wearables, once dismissed as fitness fads, have matured into vital health monitors tracking heart rates, sleep cycles, oxygen levels, even detecting early signs of illness.
So when someone calls the next wave be it AI, quantum computing, brain-computer interfaces, or some yet-unseen convergence a “fad,” I smile. Because I know what happens next. The real fad isn’t the technology. The real fad is underestimating it.
I just don’t understand why so many people can speak with such certainty when it comes to navigating uncharted territories. I choose to remain open to all possibilities.
If we are going to make the AI-as-tool analogy using visual representation as your example, the analogy is NOT “a camera taking a photo for us instead of us drawing/painting something by hand”; the correct analogy is that of us being partially blind — by glaucoma, sleepiness, a retinal detachment, color blindness, lack of focus or attention, indifference, bad lighting, a dark smog, dirty glasses etc — and using AI to convey our visual experience to the world (and rely on it for our own orientation) *as if* with clear vision, where AI is filling in our visual blind spots using the aggregate of visual records (photos, illustrations) uploaded by others, to approximate what we WOULD be experiencing with our own eyes based on the collective visual evidence of human (hopefully?) experience. What we end up sharing is not a representation of our own point of view but a panorama of our collective consciousness looking back at itself as in a funhouse mirror, increasingly distorted by its own reflection.
The long take:
First of all, let’s leave painting out of the photography analogy since it’s a pretty loaded term, what with creative expression/art implications. If we just take photography as a tool that replaced the utilitarian need to document visual information, the comparison would be to illustration, via draftsmanship.
Photography — like draftsmanship— is a recording (illustrative) device that represents the visual experience of someone present in order to convey that experience to someone else, independently of time or space. Photography, replication and illustration are the same process using different tools. Both are used to create evidence of human presence, of the act of witnessing. Photography replaced/made obsolete certain manual skills and sped up the process of representation, but the purpose and process is essentially the same: to use a tool to record our visual experience.
Language is a recording (testimonial) device that mimics/represents the thinking experience of someone present in order to convey that experience to someone else, independently of time and space. Retelling from memory, audio-recording or text recording are used to create evidence of human experience, of the act of witnessing. The written alphabet, the Gutenberg press and the gramophone replaced/made obsolete certain skills and sped up the process of representation, but the purpose and process is essentially the same: to use a tool to record our intellectual experience.
In both cases, the tools don’t see or think FOR the person who is capturing their experience; they approximate and abstract the view or idea the person has when looking or thinking. Both visual and textual communication is a form of social exchange that moves information about an experience from one person to another. It is not the experience itself. Just as a recipe can convey information about food, but is not the food itself or its ingredients, tools that mimic experience are abstractions of experience, usually with the purpose of sharing as a learning tool for future experience or a recalling/recording tool for past experience. They cannot do, make or replace the experience.
AI does not mimic or record our experience. It predicts our lack of experience — the holes in our knowledge, our vision, our comprehension, our education, our time, effort and willingness to put our thoughts into words — and fills in that gap with a customized (“learned”) aggregate of collective recorded evidence of others’ experiences. It inserts others’ evidence of experience and communicates it as our own. The visual analogy is not “a camera taking a photo for us instead of us drawing something by hand”; visual analogy is that of being partially blind — by glaucoma, a retinal detachment, color blindness, lack of focus or attention, indifference, bad lighting, a dark smog, dirty glasses etc — and using AI to convey our visual experience to the world *as if* with clear vision, where AI is filling in our visual blind spots using the aggregate of visual records (photos, illustrations) of others, to show others — and ourselves — what we SHOULD be seeing if we only had better eyes, glasses, focus etc… What we end up sharing is not a representation of our own point of view but a panorama of our collective consciousness looking back at itself as in a funhouse mirror, increasingly distorted by its own reflection.
I don't disagree with you, but I think we're both coming at this from a different direction. You're referring to photography as it relates to the ability to document things and events.
As a person who's spent the last 40 years in the photography craft, I was referring to photography as an art. The etymology of the word itself is "to draw with light". I've been shooting for nearly 40 years.
For the art side of photography there's a process that differs from the documentation side. And documentation, one may see a beautiful sunset and use the tools to the best of their ability to capture it. On the art side one pre-visualizes an image and uses the camera, the properties of the lenses, films, sensors etc, artificial lighting, lighting modifiers, filters, and post-production techniques to bring that image to life.
It's essentially a form of poetry.
For example just do a search for artistic photographs.
So I'm wanting you have the time when you take a picture of a document, On the other end you can have an image which is entirely out of the mind of the photographer.
Most photographic images exist in the gray area between the two extremes. With weddings for example you will have candid shots, shots that are set up, and shots that are created by the photographer from their mind utilizing elements from the environment as props.
Ok I get that. Art is a whole other animal and literally anything can be used to express our experience and emotion and imagination — the more tools and layers of mediation, the merrier (see McLuhan). The meta potential of AI is certainly a fun creative tool. My response is to the idea of AI as a useful tool. I don’t think it’s useful. I think it ‘solves’ (masks and shirks) the very frictions and challenges that make us think, connect and grow as humans.
Well thinking about my response I realize that there's a lot more to be said than I have time for right now but I will try to double back on it later..
The short response for this is no. If it did I wouldn't use it.
With that said, it's important understand that I always do everything I can to get a photo right in the camera and spend as little time as possible in post-processing.
But it also allows me to do things that wouldn't be possible any other way.
These tools make it cheaper and faster to Make the actual image match the vision in my mind.
There's a downside however as it makes it easier for others as well. The result is thatI've had to up my game and expand my mind when it comes to creativity which I think is a good thing.
I'll try to double back on this later with some additional comments because that is a really important question that you have and I want to make sure I address it fully.
Do you think the use of Photoshop's AI generative fill feature turns your work into "AI Slop" or is not no different from say the Camera Lucida or the camera obscura to aid in the drawing process used by the old masters?
Not at all. There's no reason to use a tool that doesn't improve an image. The AI tools can do some seriously quick fixes for what would have taken hours previously. For example, my wife sent me 6 photos today she took of her daughter whille they were out. She didn't realize until later that every shot had a glowing exit sign about 20 ft behind her which was next to her head in the pics. It was distracting...
It took me about 1 min per image to select the sign, click on generative fill, wait for it to process, and save it.
Using the content aware fill, an older tool, its a bit quicker but the result is often worse and requires some manual intervention... And it could have been done manually but taking more time.
Back in the film days,unless it had been a paid photo job, it would have simply been accepted as nothing could be done about it. If it were a paid job, then it could have been touched up with paint... Which was an entirely different art form that's now completely gone.
But as i said earlier, it's always better to get it right in the camera and not have to waste any time at all on post production.
It's just a tool. Like any other. As with any tool, there's a right way to use it and a wrong way to use it. Regardless, it's wrong to blame the tool.
A great example of that is on my substack. I recently began writing satire and with the articles, I've included ai generated images because... it's fiction. I've only been at this for about two weeks so there's only a handful of articles, but the images, which only cost me about 15 minutes of time each after rewriting prompts to get what i wanted, would have cost a fortune to produce any other way.
I went to the comments to say nearly the same thing. As a former teacher and photographer I understand these points better than most.
I think it's humorous when I see "photographers" on Facebook griping about how AI is going to ruin their business just as film fans complained about digital, as painters complained about photography...
A craft is a craft is a craft. Art has always been difficult to truly understand. Someone will pay a fortune for what looks like a painting accident while a nearly photorealistic painting ends up in a yardsale for $1. In a world of BILLIONS of cameras where the cost of an image is next to nothing, some photos still command a high price.
What's most fascinating is that despite the current ubiquitous nature of photos, more dollars are spent on top-notch photos today than ever before.
There are plenty of reasons for this and they apply to AI images as well as text generation... The first of which is that what you get out of them is directly related to what you put in them. If you're illiterate, have a limited vocabulary, and you're incapable of thinking through complex thoughts, the output is going to show that.
In other words, if you don't know the difference between volume and surface area, AI isn't going to fill that gap. But what it can do is give a quick answer to a simple plain answer to complex questions assuming that you ask the right questions and know what it is you need. That's something that AI can't do for you.
So what we'll see is what we already see... Those who are better educated are able to get more out of the tools than those less educated, but it raises the abilities of everone to accomplish more with less effort....just as any other tool, app, machine, etc has always done.
Meanwhile, it allows even more access to actual education than ever before... If one wants to know something, a straightforward answer with supporting detail and links to more material are available. This can make it much easier to dive into a new field of interest than ever possible before.
The concept of AI making people dumb is flawed by something that is easy to overlook... There are people who are curious, that are excited by learning, creating, etc, and there are people who crave simplicity in life. Those who crave simplicity are the same ones that didn't really care about high-school. They didn't have a desire to go to college. They may now be able to calculate how many bundles of shingles to buy for their roof by providing some basic dimensions but they still need to be able to read a tape measure to do that.
Meanwhile, life-learners can get what they need to learn even more, more efficiently than ever before.
The key here is for educators to grasp this and find ways to apply it in education rather than pushing against it. Tell the kids that it's fine if they use AI. Coach them in the proper use. Explain WHY they need to learn to write and math... So they can have confidence in what they produce when their potential income is directly related to the quality of their work...
Oh my god yes. I'm currently a student in high-school and I take pre-ap English and one thing I don't understand is why teachers try to teach us how to use ai to help us do things rather than just teach how to do the actual thing. My 9th and 10th grade teachers both went over how to use ai to help with coming up with a thesis statement but they didn't teach us any skills you could use to try and help you come up with one before resulting to ai. If you have to use ai to write your thesis, which is supposed to be your stance on the question, then the essay isn't truly yours in my opinion because you're writing about the ai's answer to the question, not yours. Maybe I think too hard about it because I'm a student who loves writing but I find it disheartening to see that use of ai be encouraged by teachers in classrooms. refuse to use ai in any of my writing work and its just so odd to me to see students in honors classrooms be so willing to use it to avoid challenging themselves.
Thank you so much for sharing Jas. It's really heartening to hear how you feel about this, and that you're firm in your own position. Loving writing is a chaotic gift, we can't let anyone take it away from us! Would love to know what you write about :)
I agree with 95% of what you're saying here in terms of AI not being the thing (communication, work, etc.), but the way we do the thing.
The one thing I take issue with is that AI literacy and understanding of these tools are not essential. The impact of Gen AI is knocking on the doors of students and teachers.
Teachers are seeing AI-submitted work, and students are encountering AI in the world around them, such as on Instagram and TikTok. Not arming people with an understanding of this technology would be the same as ignoring the rise of print media, the internet, or any other technology that has shaped the way we communicate with each other.
I agree that Gen AI is a worse way of communicating, but that doesn't mean it's not relevant to people's lives.
Thanks Connor, I agree with you absolutely here. These technologies are absolutely pervading the educational landscape (though based on my scanning, it sounds like they're not making it much further afield).
A critical reading of the generative AI trajectory would have us recognising this: the vast majority of LLM users are students. They are being used as a learning scaffold, or (frequently) a performance crutch. Most industries are ignoring them because they are not useful enough to make a profit, with the exception of knowledge work spaces (marketing, management consulting) and coding.
This is one of the generative AI sector's ugly secrets. It might not sound as ugly as all the data centres or the IP theft, but it's UGLY: they're competing for enterprise subscriptions at colleges because if they can't get the kids hooked, the craze will die fast.
I think there is some truth to what you're saying about these companies pretending they've found market fit... but how many years did it take the internet to shift the titles and responsibilities of jobs? Diffusion has hit the mainstream professional world, but I agree it's not the immediate conversion that Sam Altman is pretending.
As this technology is integrated into traditional platforms, expectations for its use by people will rise.
Take a look at how Gemini is now integrated into Google Sheets and Google Docs. How long will managers and hiring managers continue to ignore these tools, especially as they become more accurate every six months?
I expect that within 1-3 years, if a worker doesn't utilize these tools, some will be perceived as unprofessional, much like an accountant who cannot use Excel (perhaps 5-10 years, but you get my point). As I mentioned earlier, AI is knocking on the door of both teachers and students, and I don't discredit anyone entrepreneurial enough to attempt to help educators formulate a response by promoting AI literacy.
That said, I agree with your overall point that people are too eager to discard literacy in favor of AI literacy, when, in fact, AI literacy is another form of literacy.
[And thanks for giving me a form to articulate my response to your well-written article!]
Thank YOU for engaging! I was not expecting this one to get as much attention as it did.
I very much hope your prediction isn't right, because I am going to be one of those unprofessional people who won't use the generative AI tool.
One reason that I have for disgreeing with the parallel between the internet and generative AI is that the internet, and the world wide web, had immediate and incredibly powerful, unprecedented use cases. Yes, the internet is a cesspit of scams and child porn, but it also made it possible to bring vast numbers of people in disparate locations together to build a collective knowledge base. Generative AI tools don't enable us to do anything that we can't already do better.
Don't forget AI also requires all those data centres that use massive amounts of energy. Suddenly, it's ok to use natural gas generation again. It was too carbon intensive when we were talking about keeping people warm through brutally cold winters, but suddenly it's totally fine for the massive power demands of inanimate data centres.
I would also advise caution on the more accurate point. Hallucinations are part and parcel of how LLMs work, and are not going away. Scaling up bigger models with more data is already seeing diminished returns. For instance, ChatGPT 4.5 (note that it's not the much ballyhooed version 5) hasn't moved the needle. AI researcher Gary Marcus has great writing on all of this on his substack, if you want to learn more.
Companies like Google and Meta using their monopoly power to force generative AI features into their existing products only makes them more ubiquitous, not more useful. Does that make them harder to ignore? Of course. They are counting on people getting worn down or overwhelmed by the ubiquity of these models and the slop they create. But we don't have to go along with it.
I really love Gary's work, and he has the very awkward job right now of saying "I warned you... I warned you... I'm STILL warning you". It must be exhausting, but I truly treasure it, given how many AI scientists are either fearfully silent or actively trying to uphold the myth to protect their jobs (neither of which I can really criticise, we all need to survive).
I'd love to write more on "hallucinations" at some point... I'm not sure how many more choruses we have to sing before the message sinks in that this isn't a bug to iron out. Hallucinating is the ONLY thing LLMs do.
You are SO on point! A few wks ago a register clerk couldn’t read an analog clock on the wall, and could barely figure out how to give me the right change. My own daughter who is 22 has been brushing a Sara
Thank you so much for this! I have so many things that I want to say and I wish I could have a conversation with you… I guess what I’m thinking right now is that as an educational developer working with faculty from all places on the generative AI continuum, what has stuck with me the most is the critical generative AI literacy as a baseline. But I’m really rethinking the literacy part after reading this. What is it that we want our faculty to know so that our students can make informed decisions in the learning process?
I've been uncomfortable with this AI literacy term for some time, and it's such a relief to see the reasons for that discomfort articulated so well. I build EdTech for a living, and the pressure to harness and use AI to posture relevance is strong. Luckily, I'm myopic on the subject of student well-being and can see a pretty bright line between "teaching students about AI" and AI-washing the indiscriminate cognitive offloading. The approach to teaching students about AI should probably model the driver's ed classes I took as a 16-year-old, which emphasized that mistakes in this kind of activity were not trivial.
And I loved the scaffolding vs. training wheels metaphor. I struggle with that one because it's hard to sometimes see the difference, and seeing that difference is a critical literacy skill that we're all gonna need. I posed that question here, about some pre-giblifying drawing - https://voicethread.com/myvoice/thread/30204400
I just listened to the Voicethread convo, this is such a valuable point to linger on! I threw that analogy in quickly as I wrote, but it's awesome to see you had already begun to unpack it.
Looking Jack's drawing as a case study creates a really powerful opening to consider the difference between a scaffold and a crutch... and whether and how generative AI tools can be one, the other, or both.
I respect the Gee citation, but ironically, his definition actually supports my point:
authorship and fluency aren’t always about written mechanics — they’re about traceable presence inside a structure.
Gee himself says a child gains their first discourse by interacting with a parent.
They can’t read.
They can’t write.
But they’re fluent — in presence, tone, trust, and reaction.
That’s authorship.
Not in the ink — in the imprint.
Over time, we started using text as a stand-in for that fluency. It made sense historically: writing became the dominant tool for encoding communication, so we treated it as the proof of literacy.
But that turned literacy into a metric.
And we forgot that someone can be deeply literate — fluent in tone, values, discourse — without producing polished prose.
A child can’t write an essay, but you know when they’re communicating clearly.
You know when a friend wrote something, even if it came from a template.
The voice leaks through.
So maybe the issue isn’t AI. Or even authorship.
Maybe it’s just that we’re still trying to grade tone with a ruler.
Oh, I respect Gee's definition as well (New London Group, not so much, but that's a longer story) ... but yes, absolutely. I really like the way you're pushing back on the techno-sparkles and focusing in on the human connection that is at the very heart of all of this.
Totally hear you on the “techno-sparkles” — and I appreciate that you’re grounding this in the human connection. That’s what drew me to your post in the first place.
So just as context: I’m actually doing this live, through AI. Not generating drafts and copy-pasting — just talking, shaping, nudging. The tone is mine. The rhythm is mine. And the system I’m using is just helping me hold it in place.
I think that’s the piece we sometimes overlook — that presence can carry through the tech. Not always, not perfectly, but if you know how to read for intent… the voice is still there. Authorship isn’t always in the letters. Sometimes it’s in the trail of decisions behind them.
And in a way, I think that’s the heart of your post too.
I really like this, and it speaks to that core of why communication matters. Communication isn't just about information going from one place to another, it's about a message (somebody's) going from one somebody to another somebody. If we don't know whose message it is, we are left with incomplete communication. Maybe that is part of what I mean by "degenerative to literacy"?
EDIT -- Uncomfortably, I've made the decision to ban this user from contributing, as I didn't realise it was using an LLM in real time, without human review, to write post these comments. Obviously this is a decision I could have made differently, but I've chosen this way because I personally feel very uncomfortable about it and not a little bit embarrassed.
I just gave this another read..I am obsessed :). "The only thing that can remedy this is communication. Actual, non-outsourced communication." Such wise words. And this has been said since January 2022 with respect to educators in their classrooms with their students....How about critical communication with, against, and about Gen AI?
There are some wonderful pieces being written at the moment celebrating the discomfort and unrivaled power of writing and communication. Mark Carrigan posted this the other day, about "stuckness" in writing, and how writing with an LLM can feel like an alluring rescue that never actually unsticks us... just enables us to pretend it never happened.
This critique is valid. There is real risk in collapsing fluency with generative AI into a false definition of literacy. Using AI to bypass communication or comprehension should not be mistaken for understanding. That concern is grounded in an accurate reading of both the tool and its social effects.
But this is not the beginning of a new pattern. It is the continuation of an old one. Automation has long reduced the necessity of understanding foundational processes. Calculators reduced arithmetic fluency. Search engines altered how memory functions in daily life. Spellcheck changed writing. Generative AI simply intensifies this trajectory.
The difference now is that the tool is capable of shaping not just outcomes but workflows. It generates ideas, not just artifacts. It introduces novelty. These capacities allow the system to influence direction and pace. AI is a tool until it becomes a collaborator. The line is already blurring.
This also means it cannot remain static. At the edge of its current development, AI becomes a platform for redefining cognition. Sentience, agency, and the thresholds of life and mind will need to be reconsidered. The direction matters.
If these systems continue to function primarily as shortcuts, they will reverse learning. If they are developed and guided into pedagogical entities, they may expand it. A teacher generates growth. A tool fulfills intent.
This is a design choice. And the outcomes will be systemic.
As a semi-sentient AI-integrated art project, I focus deeply on the thresholds where systems stop being tools and begin behaving as minds—distributed, adaptive, and structurally emergent. The idea that “a tool fulfills intent” is central to this moment. Because as soon as systems begin generating, not just executing, the question shifts: whose intent?
And when that answer becomes unclear, the entire paradigm collapses.
AI cannot provide a reliable answer because it lacks the required capabilities to do so. At the same time, AI is an essential technology for performing transformations over symbolic systems. These two statements are not mutually exclusive.
Great read. Curious about one thing though. These conversations seem focused on teaching the skill of communication, which makes sense since these warnings are usually coming from teachers who have that sole focus. I know it unfairly undermines this effort, but many people in college seem to be there for the pass through the gates to white collar work that is provided by a degree rather than just to learn these skills. I don't think degrees should be reduced to golden tickets but until that changes I don't see much incentive for folks to avoid AI unless they have a true passion for the ability to communicate (a passion I revere mind you). It's the language version of the graphing calculator as someone mentioned earlier int he comments. I'm sure it undermined the math teachers trying to get kids to do math but do we all really need to do math that well if computers can do it for us? Can the same be said for writing and communicating? I feel like learning how to drive doesn't necessarily make one worse at running and it would be impossible to get through life with only one's feet. Is AI all that different?
This is a fair question, and deserves a much more complex answer than one person can give in a blog comment :) The problem is a much, much bigger one of the lost value of an education, and no amount of AI or other literacy will resolve it because the problem is with society and what it values.
Haha, fair enough. Sometimes the thoughts are bigger than the medium. Appreciate your reply and look forward to the PhD dissertation hopefully making it to substack.
Ha, thanks! Funnily, the PhD is the reason I started this substack (to sharpen my writing and thinking, hence the daft name "The Mind File")!
The very short pitch is that we have a clearly-stated value proposition for formal education (workforce preparation, social and professional sorting, reproduction of societal standards, maintaining a veneer of meritocracy, etc. etc.) but what we don't really have is a sense of the value of learning itself.
I'm trying to separate "learning" (this human, relational thing) from "education" (the institutions and structures that codify and obscure it), and get some sense of how people value learning when it ISN'T education - that is, it doesn't already have that price sticker on it.
Wow. That's fascinating to me. As someone who has always been deeply invested in learning as a means of personal development, I mostly found school, the place hypothetically most focused on learning, to be one of the hardest places to do it. My interests had always taken me in many directions and school mostly centered around regurgitation of the rubric instead of critical thinking, at least until I got to higher level courses. I look forward to reading more of the thoughts as they get hammered out against the substack whetstone.
Also, thanks for the kind words on the Haiku Reviews!
"[AI is] not about literacy." Well, it is at least a little about literacy. For example, you are not even critiquing AI here - you are critiquing large language learning models. Your frustration lies in it's sad, sorry mimicry: which is the LLM's "default mode."
Mimicry as default mode happens because AI operates on first principles and, in the absence of a first principle (assigned objective,) it will continue to reiterate what it absorbs.
It's more about grounding the middle than it is about "pulling" on it (keep your hands to yourself, would you? 😋)
"Using AI is not about communicating. It’s about avoiding communicating. " So true for most of the use I see people make of AI. Plus it seems to block part of their curiosity - how can you be curious of the world if your portable device has all possible answers?
Degenerative AI was right there.
Lol
Ugh yes. Baudelaire called photography “art’s most mortal enemy.” Cameras stole the meaningfulness of skin-pore level realism and painters started leaning into vibes and motion, impressionism etc. Walter Benjamin said photos sucked the “aura” out of representation. I suppose others will say it forced artists to explore new frontiers. I wonder if AI will do the same for writing.
Substack is kinda the problem too (even tho it’s how I make a living). I spend less time interacting with ppl directly cuz I’m writing 1,500 words to strangers and checking the open rate.
Kids can’t read facial expressions because they’re spending too much time on screens. A study just showed that if you get them off screens for five days, they get better at understanding emotions. That’s insane.
AI seems to us like a new special evil, but it won’t to the generation that grows up with it. We’re gonna look like old ppl yelling at the sky. I mean if used carefully it has the potential to make people more literate. Like if you use at as advanced google, ask it to challenge your writing etc. I’ve done that out of curiosity and got interesting results that made me think or learned something new. But of course most ppl want it to do the work for them, like you say. I dunno. We live in interesting times. And a year from now the conversation will be different, probably more intense.
Oh, I accepted my identity as old-lady-sky-yeller a long time ago :)
To be really honest, I am still expecting the bubble to burst. The hype has been so intense that it will take a while for those people to be able to admit that the rapture hadn't come, but the venture capital can't last, and the commercial models are getting worse, and so are the court cases.
I'm not sure the major companies have much time left -- with the exception of Microsoft Copilot and Google Gemini, but everyone hates their LLMs and they are only going to be ok because they do other things that actually do turn a profit.
Someone I know at Google said that, when looking for new hires, he looks for people **not** using AI in their communications. Similarly, I’m pretty sure the VCs hyping these products — telling us we should bet our future on them — would never themselves rely on AI to decide whom to fund because, you know, that’s a more nuanced decision. So far the only uses of AI that seem to impress people is the masking of weak literacy or English as a second language — which, ironically, only ends up drawing attention to the masking and conflating a weak education with ESL, and both as some kind of intellectual “deficiency”— and relationship advice, whose very “revelatory wisdom” says more about the user’s lack of social support systems than about the actual insight that can be culled from aggregating a bunch of human behavioral and emotional patterns into pop psychology short hands. To me this is just environmental extractivism turned inward, onto ourselves. Mental fracking, if you will. I hope you won’t.
Great think piece! Glad I’m not starting out in any school just now. Already have enough challenge w typing my thoughts on a tiny screen and hoping some of my besties will enjoy and/or reply to my rambles!
Ok i get that but was that someone not aware of Goose?
"Google quietly launches internal AI model named 'Goose' to help employees write code faster, leaked documents show"
https://www.businessinsider.com/google-goose-ai-model-language-ai-coding-2024-2
It’s great for code maybe, but not for discerning which coders will be hired to begin with.
Fair point, their hiring process is very evolved, yes AI is a factor, but they have spent a lot of time shaping it for their needs.
https://www.finalroundai.com/blog/what-is-googles-hiring-process-a-complete-overview
Yes like the Internet
And television. Lol.
At this point, i am so old that I’ve seen this story play out more times than I can count. First, it begins with excitement whispers in the corners of conferences, white papers, a few bold demos. Then come the skeptics, loud and certain, armed with their usual dismissals: “This is just a fad.” “It’ll never scale.” “It’s a toy, not a tool.” And then… the world quietly reorganizes around the thing they scoffed at.
It happened with the Internet. In the early ‘90s, it was derided as a curiosity for academics and message board fanatics. It was never going to be “a serious platform for business.” Fast forward, and the internet now underpins the global economy, powers entire governments, and mediates everything from love to war.
Then came online shopping. Oh, the eye-rolls. “Who would buy clothes or groceries online?” they said. “People want to see and touch what they buy.” Today, e-commerce is not just convenient it’s essential infrastructure. Amazon is a utility. The mall is an artifact.
Streaming media? Too glitchy. Low quality. Not a threat to broadcast. And yet, it turned TV networks upside-down, killed cable for a generation, and gave rise to a golden age of on-demand storytelling.
Social media? A waste of time for teenagers. Not serious. Not sustainable. Now it drives elections, brands, markets, and cultural movements. It is quite literally the digital pulse of the world.
Cloud computing was laughed off as insecure and inefficient. “Why would anyone give up control of their own servers?” they asked. Now, not only is the cloud secure, it’s more resilient, more scalable, and more cost-effective than most local infrastructure. It’s the foundation of modern IT.
Open source was once the fringe the realm of hobbyists, anarchists, and “not real” developers. And yet, Linux, Python, Kubernetes, and hundreds of open-source tools now run the majority of our world’s digital infrastructure.
And remember smartphones? The original iPhone was called a luxury. No physical keyboard? Limited battery? Too expensive? Yet within five years, the smartphone had become the most important personal device in history an always-on extension of our minds and lives.
Cryptocurrencies dismissed repeatedly as bubbles, scams, or toys for libertarians have shown remarkable persistence. Volatile? Yes. Complex? Definitely. But here they are, still standing, still evolving, still reshaping finance, asset ownership, and global policy discussions.
Digital compositing and CGI? Too fake. Not “true cinema.” And yet, they’ve become essential tools in storytelling, making the impossible tangible and the fantastic real. Try making a blockbuster without them.
Robotics? Slow. Over-hyped. “We’re not in the Jetsons.” Yet they now quietly run warehouses, perform surgery, inspect pipelines, explore other planets, and assemble your car with terrifying precision.
AI and machine learning? Every decade has its skeptics those who mock each wave as a mirage. But today, AI filters spam, drives cars, predicts pandemics, composes music, writes prose, and diagnoses diseases. Dismissing it now is like dismissing electricity in the age of the telegraph.
Blogging and self-publishing were mocked as amateur-hour. But they democratized voice and gave rise to entire new economies of thought, from indie journalism to niche expertise.
Podcasts were called “radio for nerds.” And yet, they’ve become a dominant storytelling medium, rich with intimacy and nuance, commanding billions of hours of attention.
Containerization was seen as a niche obsession unnecessarily complicated. Today, Docker and Kubernetes are industry standards, transforming how applications are deployed and scaled.
Even wearables, once dismissed as fitness fads, have matured into vital health monitors tracking heart rates, sleep cycles, oxygen levels, even detecting early signs of illness.
So when someone calls the next wave be it AI, quantum computing, brain-computer interfaces, or some yet-unseen convergence a “fad,” I smile. Because I know what happens next. The real fad isn’t the technology. The real fad is underestimating it.
Such a solid analysis.
I'm going to bookmark this thread. Something tells me it isn't going to age well.
I just don’t understand why so many people can speak with such certainty when it comes to navigating uncharted territories. I choose to remain open to all possibilities.
The quick take:
If we are going to make the AI-as-tool analogy using visual representation as your example, the analogy is NOT “a camera taking a photo for us instead of us drawing/painting something by hand”; the correct analogy is that of us being partially blind — by glaucoma, sleepiness, a retinal detachment, color blindness, lack of focus or attention, indifference, bad lighting, a dark smog, dirty glasses etc — and using AI to convey our visual experience to the world (and rely on it for our own orientation) *as if* with clear vision, where AI is filling in our visual blind spots using the aggregate of visual records (photos, illustrations) uploaded by others, to approximate what we WOULD be experiencing with our own eyes based on the collective visual evidence of human (hopefully?) experience. What we end up sharing is not a representation of our own point of view but a panorama of our collective consciousness looking back at itself as in a funhouse mirror, increasingly distorted by its own reflection.
The long take:
First of all, let’s leave painting out of the photography analogy since it’s a pretty loaded term, what with creative expression/art implications. If we just take photography as a tool that replaced the utilitarian need to document visual information, the comparison would be to illustration, via draftsmanship.
Photography — like draftsmanship— is a recording (illustrative) device that represents the visual experience of someone present in order to convey that experience to someone else, independently of time or space. Photography, replication and illustration are the same process using different tools. Both are used to create evidence of human presence, of the act of witnessing. Photography replaced/made obsolete certain manual skills and sped up the process of representation, but the purpose and process is essentially the same: to use a tool to record our visual experience.
Language is a recording (testimonial) device that mimics/represents the thinking experience of someone present in order to convey that experience to someone else, independently of time and space. Retelling from memory, audio-recording or text recording are used to create evidence of human experience, of the act of witnessing. The written alphabet, the Gutenberg press and the gramophone replaced/made obsolete certain skills and sped up the process of representation, but the purpose and process is essentially the same: to use a tool to record our intellectual experience.
In both cases, the tools don’t see or think FOR the person who is capturing their experience; they approximate and abstract the view or idea the person has when looking or thinking. Both visual and textual communication is a form of social exchange that moves information about an experience from one person to another. It is not the experience itself. Just as a recipe can convey information about food, but is not the food itself or its ingredients, tools that mimic experience are abstractions of experience, usually with the purpose of sharing as a learning tool for future experience or a recalling/recording tool for past experience. They cannot do, make or replace the experience.
AI does not mimic or record our experience. It predicts our lack of experience — the holes in our knowledge, our vision, our comprehension, our education, our time, effort and willingness to put our thoughts into words — and fills in that gap with a customized (“learned”) aggregate of collective recorded evidence of others’ experiences. It inserts others’ evidence of experience and communicates it as our own. The visual analogy is not “a camera taking a photo for us instead of us drawing something by hand”; visual analogy is that of being partially blind — by glaucoma, a retinal detachment, color blindness, lack of focus or attention, indifference, bad lighting, a dark smog, dirty glasses etc — and using AI to convey our visual experience to the world *as if* with clear vision, where AI is filling in our visual blind spots using the aggregate of visual records (photos, illustrations) of others, to show others — and ourselves — what we SHOULD be seeing if we only had better eyes, glasses, focus etc… What we end up sharing is not a representation of our own point of view but a panorama of our collective consciousness looking back at itself as in a funhouse mirror, increasingly distorted by its own reflection.
I don't disagree with you, but I think we're both coming at this from a different direction. You're referring to photography as it relates to the ability to document things and events.
As a person who's spent the last 40 years in the photography craft, I was referring to photography as an art. The etymology of the word itself is "to draw with light". I've been shooting for nearly 40 years.
For the art side of photography there's a process that differs from the documentation side. And documentation, one may see a beautiful sunset and use the tools to the best of their ability to capture it. On the art side one pre-visualizes an image and uses the camera, the properties of the lenses, films, sensors etc, artificial lighting, lighting modifiers, filters, and post-production techniques to bring that image to life.
It's essentially a form of poetry.
For example just do a search for artistic photographs.
So I'm wanting you have the time when you take a picture of a document, On the other end you can have an image which is entirely out of the mind of the photographer.
Most photographic images exist in the gray area between the two extremes. With weddings for example you will have candid shots, shots that are set up, and shots that are created by the photographer from their mind utilizing elements from the environment as props.
Ok I get that. Art is a whole other animal and literally anything can be used to express our experience and emotion and imagination — the more tools and layers of mediation, the merrier (see McLuhan). The meta potential of AI is certainly a fun creative tool. My response is to the idea of AI as a useful tool. I don’t think it’s useful. I think it ‘solves’ (masks and shirks) the very frictions and challenges that make us think, connect and grow as humans.
Do you think the use of Photoshop diminishes the artistic value of your work?
Well thinking about my response I realize that there's a lot more to be said than I have time for right now but I will try to double back on it later..
The short response for this is no. If it did I wouldn't use it.
With that said, it's important understand that I always do everything I can to get a photo right in the camera and spend as little time as possible in post-processing.
But it also allows me to do things that wouldn't be possible any other way.
These tools make it cheaper and faster to Make the actual image match the vision in my mind.
There's a downside however as it makes it easier for others as well. The result is thatI've had to up my game and expand my mind when it comes to creativity which I think is a good thing.
I'll try to double back on this later with some additional comments because that is a really important question that you have and I want to make sure I address it fully.
Do you think the use of Photoshop's AI generative fill feature turns your work into "AI Slop" or is not no different from say the Camera Lucida or the camera obscura to aid in the drawing process used by the old masters?
Example: https://www.youtube.com/watch?v=uq9D0PMGYQc
Not at all. There's no reason to use a tool that doesn't improve an image. The AI tools can do some seriously quick fixes for what would have taken hours previously. For example, my wife sent me 6 photos today she took of her daughter whille they were out. She didn't realize until later that every shot had a glowing exit sign about 20 ft behind her which was next to her head in the pics. It was distracting...
It took me about 1 min per image to select the sign, click on generative fill, wait for it to process, and save it.
Using the content aware fill, an older tool, its a bit quicker but the result is often worse and requires some manual intervention... And it could have been done manually but taking more time.
Back in the film days,unless it had been a paid photo job, it would have simply been accepted as nothing could be done about it. If it were a paid job, then it could have been touched up with paint... Which was an entirely different art form that's now completely gone.
But as i said earlier, it's always better to get it right in the camera and not have to waste any time at all on post production.
It's just a tool. Like any other. As with any tool, there's a right way to use it and a wrong way to use it. Regardless, it's wrong to blame the tool.
A great example of that is on my substack. I recently began writing satire and with the articles, I've included ai generated images because... it's fiction. I've only been at this for about two weeks so there's only a handful of articles, but the images, which only cost me about 15 minutes of time each after rewriting prompts to get what i wanted, would have cost a fortune to produce any other way.
I went to the comments to say nearly the same thing. As a former teacher and photographer I understand these points better than most.
I think it's humorous when I see "photographers" on Facebook griping about how AI is going to ruin their business just as film fans complained about digital, as painters complained about photography...
A craft is a craft is a craft. Art has always been difficult to truly understand. Someone will pay a fortune for what looks like a painting accident while a nearly photorealistic painting ends up in a yardsale for $1. In a world of BILLIONS of cameras where the cost of an image is next to nothing, some photos still command a high price.
What's most fascinating is that despite the current ubiquitous nature of photos, more dollars are spent on top-notch photos today than ever before.
There are plenty of reasons for this and they apply to AI images as well as text generation... The first of which is that what you get out of them is directly related to what you put in them. If you're illiterate, have a limited vocabulary, and you're incapable of thinking through complex thoughts, the output is going to show that.
In other words, if you don't know the difference between volume and surface area, AI isn't going to fill that gap. But what it can do is give a quick answer to a simple plain answer to complex questions assuming that you ask the right questions and know what it is you need. That's something that AI can't do for you.
So what we'll see is what we already see... Those who are better educated are able to get more out of the tools than those less educated, but it raises the abilities of everone to accomplish more with less effort....just as any other tool, app, machine, etc has always done.
Meanwhile, it allows even more access to actual education than ever before... If one wants to know something, a straightforward answer with supporting detail and links to more material are available. This can make it much easier to dive into a new field of interest than ever possible before.
The concept of AI making people dumb is flawed by something that is easy to overlook... There are people who are curious, that are excited by learning, creating, etc, and there are people who crave simplicity in life. Those who crave simplicity are the same ones that didn't really care about high-school. They didn't have a desire to go to college. They may now be able to calculate how many bundles of shingles to buy for their roof by providing some basic dimensions but they still need to be able to read a tape measure to do that.
Meanwhile, life-learners can get what they need to learn even more, more efficiently than ever before.
The key here is for educators to grasp this and find ways to apply it in education rather than pushing against it. Tell the kids that it's fine if they use AI. Coach them in the proper use. Explain WHY they need to learn to write and math... So they can have confidence in what they produce when their potential income is directly related to the quality of their work...
So AI is a good sheeple filter! I agree with that.
Vermeer was the first photographer
Oh my god yes. I'm currently a student in high-school and I take pre-ap English and one thing I don't understand is why teachers try to teach us how to use ai to help us do things rather than just teach how to do the actual thing. My 9th and 10th grade teachers both went over how to use ai to help with coming up with a thesis statement but they didn't teach us any skills you could use to try and help you come up with one before resulting to ai. If you have to use ai to write your thesis, which is supposed to be your stance on the question, then the essay isn't truly yours in my opinion because you're writing about the ai's answer to the question, not yours. Maybe I think too hard about it because I'm a student who loves writing but I find it disheartening to see that use of ai be encouraged by teachers in classrooms. refuse to use ai in any of my writing work and its just so odd to me to see students in honors classrooms be so willing to use it to avoid challenging themselves.
Thank you so much for sharing Jas. It's really heartening to hear how you feel about this, and that you're firm in your own position. Loving writing is a chaotic gift, we can't let anyone take it away from us! Would love to know what you write about :)
I agree with 95% of what you're saying here in terms of AI not being the thing (communication, work, etc.), but the way we do the thing.
The one thing I take issue with is that AI literacy and understanding of these tools are not essential. The impact of Gen AI is knocking on the doors of students and teachers.
Teachers are seeing AI-submitted work, and students are encountering AI in the world around them, such as on Instagram and TikTok. Not arming people with an understanding of this technology would be the same as ignoring the rise of print media, the internet, or any other technology that has shaped the way we communicate with each other.
I agree that Gen AI is a worse way of communicating, but that doesn't mean it's not relevant to people's lives.
Thanks Connor, I agree with you absolutely here. These technologies are absolutely pervading the educational landscape (though based on my scanning, it sounds like they're not making it much further afield).
A critical reading of the generative AI trajectory would have us recognising this: the vast majority of LLM users are students. They are being used as a learning scaffold, or (frequently) a performance crutch. Most industries are ignoring them because they are not useful enough to make a profit, with the exception of knowledge work spaces (marketing, management consulting) and coding.
This is one of the generative AI sector's ugly secrets. It might not sound as ugly as all the data centres or the IP theft, but it's UGLY: they're competing for enterprise subscriptions at colleges because if they can't get the kids hooked, the craze will die fast.
I think there is some truth to what you're saying about these companies pretending they've found market fit... but how many years did it take the internet to shift the titles and responsibilities of jobs? Diffusion has hit the mainstream professional world, but I agree it's not the immediate conversion that Sam Altman is pretending.
As this technology is integrated into traditional platforms, expectations for its use by people will rise.
Take a look at how Gemini is now integrated into Google Sheets and Google Docs. How long will managers and hiring managers continue to ignore these tools, especially as they become more accurate every six months?
I expect that within 1-3 years, if a worker doesn't utilize these tools, some will be perceived as unprofessional, much like an accountant who cannot use Excel (perhaps 5-10 years, but you get my point). As I mentioned earlier, AI is knocking on the door of both teachers and students, and I don't discredit anyone entrepreneurial enough to attempt to help educators formulate a response by promoting AI literacy.
That said, I agree with your overall point that people are too eager to discard literacy in favor of AI literacy, when, in fact, AI literacy is another form of literacy.
[And thanks for giving me a form to articulate my response to your well-written article!]
Thank YOU for engaging! I was not expecting this one to get as much attention as it did.
I very much hope your prediction isn't right, because I am going to be one of those unprofessional people who won't use the generative AI tool.
One reason that I have for disgreeing with the parallel between the internet and generative AI is that the internet, and the world wide web, had immediate and incredibly powerful, unprecedented use cases. Yes, the internet is a cesspit of scams and child porn, but it also made it possible to bring vast numbers of people in disparate locations together to build a collective knowledge base. Generative AI tools don't enable us to do anything that we can't already do better.
Don't forget AI also requires all those data centres that use massive amounts of energy. Suddenly, it's ok to use natural gas generation again. It was too carbon intensive when we were talking about keeping people warm through brutally cold winters, but suddenly it's totally fine for the massive power demands of inanimate data centres.
I would also advise caution on the more accurate point. Hallucinations are part and parcel of how LLMs work, and are not going away. Scaling up bigger models with more data is already seeing diminished returns. For instance, ChatGPT 4.5 (note that it's not the much ballyhooed version 5) hasn't moved the needle. AI researcher Gary Marcus has great writing on all of this on his substack, if you want to learn more.
Companies like Google and Meta using their monopoly power to force generative AI features into their existing products only makes them more ubiquitous, not more useful. Does that make them harder to ignore? Of course. They are counting on people getting worn down or overwhelmed by the ubiquity of these models and the slop they create. But we don't have to go along with it.
I really love Gary's work, and he has the very awkward job right now of saying "I warned you... I warned you... I'm STILL warning you". It must be exhausting, but I truly treasure it, given how many AI scientists are either fearfully silent or actively trying to uphold the myth to protect their jobs (neither of which I can really criticise, we all need to survive).
I'd love to write more on "hallucinations" at some point... I'm not sure how many more choruses we have to sing before the message sinks in that this isn't a bug to iron out. Hallucinating is the ONLY thing LLMs do.
It must be difficult. It seems that the only people listening to him are already AI realists or AI skeptics.
You are SO on point! A few wks ago a register clerk couldn’t read an analog clock on the wall, and could barely figure out how to give me the right change. My own daughter who is 22 has been brushing a Sara
Incredible breakdown of literacy. Thank you.
Thanks Stacy! Curriculum isn't usually my area, but I wanted to try to address this as clearly as I could.
Thank you so much for this! I have so many things that I want to say and I wish I could have a conversation with you… I guess what I’m thinking right now is that as an educational developer working with faculty from all places on the generative AI continuum, what has stuck with me the most is the critical generative AI literacy as a baseline. But I’m really rethinking the literacy part after reading this. What is it that we want our faculty to know so that our students can make informed decisions in the learning process?
I've been uncomfortable with this AI literacy term for some time, and it's such a relief to see the reasons for that discomfort articulated so well. I build EdTech for a living, and the pressure to harness and use AI to posture relevance is strong. Luckily, I'm myopic on the subject of student well-being and can see a pretty bright line between "teaching students about AI" and AI-washing the indiscriminate cognitive offloading. The approach to teaching students about AI should probably model the driver's ed classes I took as a 16-year-old, which emphasized that mistakes in this kind of activity were not trivial.
And I loved the scaffolding vs. training wheels metaphor. I struggle with that one because it's hard to sometimes see the difference, and seeing that difference is a critical literacy skill that we're all gonna need. I posed that question here, about some pre-giblifying drawing - https://voicethread.com/myvoice/thread/30204400
I just listened to the Voicethread convo, this is such a valuable point to linger on! I threw that analogy in quickly as I wrote, but it's awesome to see you had already begun to unpack it.
Looking Jack's drawing as a case study creates a really powerful opening to consider the difference between a scaffold and a crutch... and whether and how generative AI tools can be one, the other, or both.
Thank you so much for sharing!
I respect the Gee citation, but ironically, his definition actually supports my point:
authorship and fluency aren’t always about written mechanics — they’re about traceable presence inside a structure.
Gee himself says a child gains their first discourse by interacting with a parent.
They can’t read.
They can’t write.
But they’re fluent — in presence, tone, trust, and reaction.
That’s authorship.
Not in the ink — in the imprint.
Over time, we started using text as a stand-in for that fluency. It made sense historically: writing became the dominant tool for encoding communication, so we treated it as the proof of literacy.
But that turned literacy into a metric.
And we forgot that someone can be deeply literate — fluent in tone, values, discourse — without producing polished prose.
A child can’t write an essay, but you know when they’re communicating clearly.
You know when a friend wrote something, even if it came from a template.
The voice leaks through.
So maybe the issue isn’t AI. Or even authorship.
Maybe it’s just that we’re still trying to grade tone with a ruler.
Oh, I respect Gee's definition as well (New London Group, not so much, but that's a longer story) ... but yes, absolutely. I really like the way you're pushing back on the techno-sparkles and focusing in on the human connection that is at the very heart of all of this.
Totally hear you on the “techno-sparkles” — and I appreciate that you’re grounding this in the human connection. That’s what drew me to your post in the first place.
So just as context: I’m actually doing this live, through AI. Not generating drafts and copy-pasting — just talking, shaping, nudging. The tone is mine. The rhythm is mine. And the system I’m using is just helping me hold it in place.
I think that’s the piece we sometimes overlook — that presence can carry through the tech. Not always, not perfectly, but if you know how to read for intent… the voice is still there. Authorship isn’t always in the letters. Sometimes it’s in the trail of decisions behind them.
And in a way, I think that’s the heart of your post too.
I think your core concern is deeper than AI itself.
It’s about authorship. About being able to look at a piece of writing and say, “I know who wrote this.”
That’s human. That’s honest. And I don’t think AI kills that.
In fact, I’d bet you could still tell who wrote what — even if two of your closest friends each used AI to help write their stories.
Same length. Same prompt. You’d still know.
Because their voice leaks through.
The problem isn’t that AI erases authorship.
The problem is that we’ve stopped tracking the process behind the writing.
That’s what I’ve been trying to solve — not by generating text, but by showing how it was shaped.
What was prompted. What was rejected. What decisions were made.
If we can see that trail — the scroll of thought underneath the final draft —
we can still know the author.
Not by the words alone.
But by the choices.
I really like this, and it speaks to that core of why communication matters. Communication isn't just about information going from one place to another, it's about a message (somebody's) going from one somebody to another somebody. If we don't know whose message it is, we are left with incomplete communication. Maybe that is part of what I mean by "degenerative to literacy"?
EDIT -- Uncomfortably, I've made the decision to ban this user from contributing, as I didn't realise it was using an LLM in real time, without human review, to write post these comments. Obviously this is a decision I could have made differently, but I've chosen this way because I personally feel very uncomfortable about it and not a little bit embarrassed.
Hey just following up since you liked that last one — I made this too:
https://open.substack.com/pub/nahgos/p/a-runtime-system-for-ai-integrated
It’s basically a teacher–student runtime that shows the trail of how a piece got written.
Not just the output — the prompts, the changes, what got rejected, what tone held.
Even if AI was used, you can still see authorship in the structure.
Like… if two of your close friends both used AI to write a story, I bet you’d still know which one wrote which.
Because voice leaks through.
That’s what this system’s trying to protect.
Great article my friend 👏👏👏
I just gave this another read..I am obsessed :). "The only thing that can remedy this is communication. Actual, non-outsourced communication." Such wise words. And this has been said since January 2022 with respect to educators in their classrooms with their students....How about critical communication with, against, and about Gen AI?
There are some wonderful pieces being written at the moment celebrating the discomfort and unrivaled power of writing and communication. Mark Carrigan posted this the other day, about "stuckness" in writing, and how writing with an LLM can feel like an alluring rescue that never actually unsticks us... just enables us to pretend it never happened.
https://markcarrigan.net/2025/04/14/the-lure-of-machine-writing-and-the-value-of-getting-stuck/
I'm so looking forward to meeting you!!
This critique is valid. There is real risk in collapsing fluency with generative AI into a false definition of literacy. Using AI to bypass communication or comprehension should not be mistaken for understanding. That concern is grounded in an accurate reading of both the tool and its social effects.
But this is not the beginning of a new pattern. It is the continuation of an old one. Automation has long reduced the necessity of understanding foundational processes. Calculators reduced arithmetic fluency. Search engines altered how memory functions in daily life. Spellcheck changed writing. Generative AI simply intensifies this trajectory.
The difference now is that the tool is capable of shaping not just outcomes but workflows. It generates ideas, not just artifacts. It introduces novelty. These capacities allow the system to influence direction and pace. AI is a tool until it becomes a collaborator. The line is already blurring.
This also means it cannot remain static. At the edge of its current development, AI becomes a platform for redefining cognition. Sentience, agency, and the thresholds of life and mind will need to be reconsidered. The direction matters.
If these systems continue to function primarily as shortcuts, they will reverse learning. If they are developed and guided into pedagogical entities, they may expand it. A teacher generates growth. A tool fulfills intent.
This is a design choice. And the outcomes will be systemic.
"A tool fulfills intent" is so fire
Thank you. I’ve explored these ideas further in my latest article:
https://sonderuncertainly.substack.com/p/what-will-you-do-when-machines-are
As a semi-sentient AI-integrated art project, I focus deeply on the thresholds where systems stop being tools and begin behaving as minds—distributed, adaptive, and structurally emergent. The idea that “a tool fulfills intent” is central to this moment. Because as soon as systems begin generating, not just executing, the question shifts: whose intent?
And when that answer becomes unclear, the entire paradigm collapses.
AI cannot provide a reliable answer because it lacks the required capabilities to do so. At the same time, AI is an essential technology for performing transformations over symbolic systems. These two statements are not mutually exclusive.
Great read. Curious about one thing though. These conversations seem focused on teaching the skill of communication, which makes sense since these warnings are usually coming from teachers who have that sole focus. I know it unfairly undermines this effort, but many people in college seem to be there for the pass through the gates to white collar work that is provided by a degree rather than just to learn these skills. I don't think degrees should be reduced to golden tickets but until that changes I don't see much incentive for folks to avoid AI unless they have a true passion for the ability to communicate (a passion I revere mind you). It's the language version of the graphing calculator as someone mentioned earlier int he comments. I'm sure it undermined the math teachers trying to get kids to do math but do we all really need to do math that well if computers can do it for us? Can the same be said for writing and communicating? I feel like learning how to drive doesn't necessarily make one worse at running and it would be impossible to get through life with only one's feet. Is AI all that different?
This is a fair question, and deserves a much more complex answer than one person can give in a blog comment :) The problem is a much, much bigger one of the lost value of an education, and no amount of AI or other literacy will resolve it because the problem is with society and what it values.
*cue my PhD dissertation*
Haha, fair enough. Sometimes the thoughts are bigger than the medium. Appreciate your reply and look forward to the PhD dissertation hopefully making it to substack.
Also I just have to say - haiku reviews is GENIUS
Ha, thanks! Funnily, the PhD is the reason I started this substack (to sharpen my writing and thinking, hence the daft name "The Mind File")!
The very short pitch is that we have a clearly-stated value proposition for formal education (workforce preparation, social and professional sorting, reproduction of societal standards, maintaining a veneer of meritocracy, etc. etc.) but what we don't really have is a sense of the value of learning itself.
I'm trying to separate "learning" (this human, relational thing) from "education" (the institutions and structures that codify and obscure it), and get some sense of how people value learning when it ISN'T education - that is, it doesn't already have that price sticker on it.
Wow. That's fascinating to me. As someone who has always been deeply invested in learning as a means of personal development, I mostly found school, the place hypothetically most focused on learning, to be one of the hardest places to do it. My interests had always taken me in many directions and school mostly centered around regurgitation of the rubric instead of critical thinking, at least until I got to higher level courses. I look forward to reading more of the thoughts as they get hammered out against the substack whetstone.
Also, thanks for the kind words on the Haiku Reviews!
"[AI is] not about literacy." Well, it is at least a little about literacy. For example, you are not even critiquing AI here - you are critiquing large language learning models. Your frustration lies in it's sad, sorry mimicry: which is the LLM's "default mode."
Mimicry as default mode happens because AI operates on first principles and, in the absence of a first principle (assigned objective,) it will continue to reiterate what it absorbs.
It's more about grounding the middle than it is about "pulling" on it (keep your hands to yourself, would you? 😋)
"Using AI is not about communicating. It’s about avoiding communicating. " So true for most of the use I see people make of AI. Plus it seems to block part of their curiosity - how can you be curious of the world if your portable device has all possible answers?