Thanks for including my piece and for taking a look at these AI checkers! Super important given how widely they’re being promoted. Fwiw, my writing is always “100% human” (as anyone familiar with my POV on the subject will tell you) so the "mixed" result is likely a false positive….which is a pretty good illustration of why I'm skeptical of these tools in the first place 😵💫 Appreciate the mention!
Thanks for writing - was just ranting about this very thing to a friend last week.
And it gets worse, just this morning I get a WhatsApp message from a friend I text with regularly, who wanted me to get a family plan nyt subscription with him. I ask him why he wants to pay for nyt, and dude sends me a wall of GPT generated text. To be fair he did admit it in a follow up, but it was just so gross to have to parse AI text in a personal chat. Nowhere is safe it seems. Sorry for the long rant
People have been outsourcing their thinking and their words for a long time. AI is just a more efficient means of doing so.
Is it particularly sloppy and dishonest in the presentation of one's alleged writing? Of course. But just as AI sources from the "Public" that is reddit and the larger internet, people have been allowing the Public to think and speak for them forever, as Kierkegaard pointed out in detail.
All this is to say that most people were unoriginal parrots even before they had LLMs to do the parroting for them.
This is why I vowed early never to use AI in my writing. I even went so far as to write my second book of poetry offline, editing on a computer that does not go online. The text has never seen the internet.
When the first US Navy destroyers arrived in Queenstown, Ireland in 1917 to join the allied war against the Kaiser’s U-boats, the American commander reported his arrival to the British admiral at his headquarters. The admiral could not see him immediately because he was transcribing telegrams from his combat ships into longhand for his war report to the Admiralty in London. He was doing this because that is what Lord Admiral Nelson did when he was fighting Napoleon’s fleets. In 1805.
My mother worked on the B 52 program when it was being designed and the prototype was under construction. She received formulas and tables of numbers from the engineers and gave them back her calculations. Her job title was “Computer.” She computed.
That function is still in existence today, embedded in the various CAD/CAM systems engineers use to design modern aircraft.
Are today’s designs “less human” than they were in 1951 because the embedded computer is now digital?
When I was in high school and college we were prohibited from using electronic calculators in class. They would inhibit our ability to understand mathematics. That is what we were told.
When I was writing my Master’s Thesis (on heuristic AI, oddly) I was not allowed to submit a version written on a word processor. I had to type it on a typewriter or hire a professional secretary to generate the final, paper product. Typed on a typewriter. The physical act of writing was seen an important part of the academic process.
Nelson embedded the “Nelson touch” for decisive victory at sea into future admirals through longhand reporting instead of relying on telegraph slop.
My mother saved engineers from “computing slop.”
My high school teachers saved me from “mathematics slop.”
My thesis advisors saved me from “word processor slop.”
Dr. Richard Hamming, a Turing prize winner (and one of my professors) wrote “"Change does not mean progress, but progress requires change."
I offer that today’s slop is tomorrow’s “can you believe they actually did that?”
"When I was in high school and college we were prohibited from using electronic calculators in class. They would inhibit our ability to understand mathematics. That is what we were told."
Your teachers were correct.
Writing is different than mathematics as it relies on emotional rhetoric and appeals to common experience, something mathematics, as a purely abstract art, does not. AI tries to mimic it, but is fraudulent for the same reason having someone else write a letter to your wife and passing it as your own is fraudulent.
When I read academic writing from a century ago, I see references to Latin and Greek languages and texts. Untranslated, because everyone then knew those languages and the texts to which they referred. Toynbee is a recent example for me in my reading and research.
I lived with my maternal grandmother for a year in high school. She would quote Cicero and Tennyson at length from memory. That’s what she learned in school as a child.
Am I uneducated because I’m not literate in those? Is my writing inadequate because of that?
It does. What was authentic a century ago is not the same as authentic today. Today’s human writing would not be considered authentic a century ago.
I offer that what you define as authentic today will not be considered authentic in the future.
That is Hamming’s progress. You hit on it exactly with your opening example about craftsmanship. Today’s is not the same as yesterday’s. Where I differ from you is that I assert that today’s craftsmanship, and today’s AI writing are just fine; they have progressed. They are not the same as before but that that doesn’t make them “bad.”
Just like the format for Royal Navy reports, computers, and calculators. Different, yes. Bad, no. Better, yes.
Not having to hand tool metalwork because it is produced correctly is not bad. Not having to mash numbers to calculate the roots of a polynomial is not bad. Not having to labor for a week to create a 3000 word essay is not bad. All of those things represent technical, social, and cultural progress. They are all good.
The practitioners from the prior eras would probably not agree. “I had to learn Greek and Latin and memorize hundreds of lines of ancient prose; who are you to say you don’t have to?”
"It does. What was authentic a century ago is not the same as authentic today. Today’s human writing would not be considered authentic a century ago."
What do you mean by this? A love letter in the civil war era is just as authentic as a modern love letter. What has never been considered authentic is outsourcing the work to someone else.
"Claude, write my wife a love letter telling her I miss her" is far different than putting your emotions into words. If you can't see that I don't think I can help you.
I'm using an engineering history book written in 1898 for a project. Every chapter starts with a poem, a bible verse, a classic quotation, or other philosophical item relevant to the topic to be discussed in that chapter. And every chapter starts with a lengthy summary of the chapter (essentially a list of the headings). There are 40 chapters. It is a very technical history: pressures, temperatures, design drawings, operating instructions, etc. Not literature. Nothing like that is written today in that format or style. If I gave that author a modern engineering history, he would reject it as inadequate, incomplete, and stylistically unacceptable. Slop. If I input his text into a modern LLM, I would get a similar response, in the opposite vein. Most importantly, if I submitted that "authentic" 1898 format to a modern peer-reviewed journal, it would be rejected, regardless of the merit of my content. Slop.
I'll accept that love letters don't change much over the generations. The bulk of the writing I see is technical, programmatic, or policy-related for work, education, or advocacy. It is VERY attuned to the current day, practices, styles, and cultural mores. And those are constantly changing. LLM AI does a superb job of keeping up with those changes and that level of "progress," so I don't see it as slop. I see it as a valuable tool for meeting my daily performance requirements in my writing.
I'll concede the literature & personal letter slop points, mostly due to a lack of knowledge on my part. I do know that Sailors, soldiers, and prisoners have other people write their personal letters all the time. I hope that is not shocking to you.
A key point: LLM AI is not a person. It is a computer program. It is a tool, a machine, an inanimate object like a hammer, a leatherman, a calculator, or a spreadsheet. Having an LLM do something for me is using a hammer to knock off some rust, or a screwdriver to install a fastener. There is no moral or ethical conflict in the machine except in how I use it ( I can kill somebody with a hammer as easily as I drive a nail; the morality is mine, not the hammer’s).
For me, LLM AI text is not slop at all; like a calculator or a computer, it is an essential tool for getting my work done well, effectively, and efficiently. I do have to be able to use the tool; that can be a challenge for some people, especially at first, while they're learning it. As with any new tool.
I liked your post. Thank you for writing it. It surfaces an issue that accompanies all changes in technology, and has for all of human history. It’s good to get the topic on the table. I’ve read a thousand such posts & laments covering all the recorded millennia of human history and for all human technologies. Unfortunately, like your predecessors, I think you’re pushing a rope, so to speak. Dr. Hamming got it right. All progress requires change.
When I was taking our control systems class in graduate school, our grades were determined by the quality of the calculators we had purchased. The key activity in control systems is finding the roots of n-dimensional polynomials. Our ability to do that was 100% dependent on the quality (i.e. price) of our calculators. Our professor gave us brand and model numbers of the calculators needed to get a passing grade.
I'm not above naming one: NPOV checks as 100% AI on several articles. They're a Wikipedia watchdog, making their decision to employ AI rather involuted: We now have AI slop used as an investigative tool of AI slop. That's a lot of disrespect for Wikipedia. Maybe they're doing it right.
Heh, great minds think alike! I can’t stand any of it, in the classroom or in publications.
Funny story. I remember reading an obvious AI-counterfeit essay in Acton for whom I used to occasionally write, and I let the editor, Anthony Sacramone, know about it. I was nice about it and honestly thought he’d take my point in the proper spirit. Nope! He got pissed and told me that I couldn’t write for them anymore. I even apologized afterward for upsetting him (even though I was totally right and had a bunch of scans to prove it). But I’m still blackballed. So yeah, screw them.
I don’t think it’s all AI slop at Acton, but it’s definitely happening with some of their lesser known “young” writers. And yes, I definitely encountered pretentiousness and anti-populist bias from the editor. This is pretty common with the self-serious boutique publications like Acton.
I literally stopped using Substack for a few months because I was inundated with AI slop. I would love to get your thoughts on the project I’m starting! Take a look at my page to get a sense for what I’m thinking
I decided to test it with one of mine(I’m still skeptical and am on the free version) and it came back 100% human, so that is a mark in its favor so far.
Let me clear something up, because being filed next to crypto spam and get-rich-quick slop is a stretch, even for the internet.
I'm a PA-C with a PhD and more than twenty years in clinical medicine, ten of them in an emergency room. Every lab range on my Substack is mine. Every protocol is mine. Every opinion is mine, built from my own training and the years I've spent reading studies most people scroll right past.
Yes, I use AI. Here's the part the detector can't tell you. There's a difference between using AI and training it, and that difference is the whole ballgame.
Using AI is asking the machine what it thinks. It answers with the average of the internet, and the average of the internet is how you get slop. Confident, tidy, and often wrong about your health.
Training it is the opposite. I've fed it my standards, my ranges, the studies I trust, two decades of clinical reasoning, and now it checks my work against me instead of against Google. It flags when a number looks off so I catch it before 10,000 people read it. It tells me when a study doesn't say what I think it says. It cleans my grammar on the way out.
There's a name you already trust for that. Spellcheck. A spellchecker flags "teh" because someone loaded a dictionary into it, and nobody has ever called that cheating. I loaded twenty years of medicine into the tool instead of a dictionary. Same idea, more horsepower.
So no, the machine does not write for me. It helps me answer hundreds of comments from people asking for real help, because there is one of me and thousands of them, and I'd rather reach all of them well than a few of them slowly.
You brought up your grandfather carving birds by hand, and that you treasure them. I get that. I spent eleven years inside the cancer system supporting my daughter, and in that time I built a framework that now helps thousands of women understand their own bodies for the first time. Using a power tool to sand the edges doesn't make the bird less his. The hand and the mind are what make it real, and those are mine.
Here's the irony I keep turning over. You ran my Substack through a tool to decide I lean on tools. The difference is I read my work before I publish it. You didn't read mine before you judged it.
If connection is your goal why have your comment be written by an ai? Thats what it sounds like. If you did write it and want connection with human beings why do you write like an ai?
Be very careful about your accusations and beliefs. You may have a tool that you think is smart, and is giving you the truth, and that this truth is yours to crusade with, but Jordan Dann’s writing is all her own. Her thoughts, her research, her years of professional practice and development. How do I know? Because I watch her week after week painstakingly crafting what she shares with her audience for hours upon hours in her office. I bring her the coffee because she is so engrossed in her writing. I walk the dog so she can spend more time revising. We discuss her ideas on our couch. I see the process. Yes, AI is helpful in grammar and spelling and in other useful ways. But please do not trample on her work in some finger-pointing “Aha!” fashion. It is misguided.
My dilemma on utoob is that sometimes there is something I didn't know about, military history for example. I start the video. 30 seconds in it is blatantly obvious AI, yet the subject matter is something I want to know about, but I am angry at the content "creator" for using AI. Do I click away immediately, or stick around to learn something? Should I bother to complain about AI use in the comments?
I *really* do not want AI crap on Substack, that's not what I'm here for.
If it's any consolation, I'm convinced that the proliferation of AI is a fad. While it can be a useful tool and it will continue to be used in some areas, people hate it. I think we're on the left side of a Gaussian curve. Usage will increase, peak, and then fade. Grok is a money pit for X. The data center craze is hated and cannot be supported by our electrical infrastructure. For now we can complain, mock, avoid, and protest until the wave inevitably passes.
I get it - the AI fingerprints are really making Substack quite grimy. But I think your approach is wrong. You berate those who outsource their writing/thinking to machines but you are outsourcing your sense for quality, your taste, to Pangram. If the human connection is what you pine for, shouldn't you be able to judge whether that connection exists just by reading the posts? Why give a machine such sway over your judgement?
The AI checker tools can't tell whether a human writer is involved for the same reason that LLMs can't produce writing that feels human - because neither has any understanding of what it is to be human. You've probably outed some hard working authors who just write in an old-school, Strunk's-Elements-of-Style manner, which LLMs are expert at imitating, rather than in a more personal, idiosyncratic way, which LLMs can't do.
"If the human connection is what you pine for, shouldn't you be able to judge whether that connection exists just by reading the posts? Why give a machine such sway over your judgement?"
If one is on alert for AI writing, it often takes a decent amount of reading to discern AI writing. If you are on alert for AI writing, it is harder to enjoy the article. Either way is non-optimal, and the checker is the best way to ensure my time is not being wasted and I'm reading a human author.
"You've probably outed some hard working authors who just write in an old-school, Strunk's-Elements-of-Style manner, which LLMs are expert at imitating, rather than in a more personal, idiosyncratic way, which LLMs can't do."
Great article, and very timely. I do think you should publicly expose the accounts in our sphere who are predominantly using AI, if only as a public safety measure. People need to know what they’re reading, I for one would love to know whether the stuff I read is actually written by a person or not.
Thanks for including my piece and for taking a look at these AI checkers! Super important given how widely they’re being promoted. Fwiw, my writing is always “100% human” (as anyone familiar with my POV on the subject will tell you) so the "mixed" result is likely a false positive….which is a pretty good illustration of why I'm skeptical of these tools in the first place 😵💫 Appreciate the mention!
I updated it with your response. Thanks.
My pleasure!
Thanks for writing - was just ranting about this very thing to a friend last week.
And it gets worse, just this morning I get a WhatsApp message from a friend I text with regularly, who wanted me to get a family plan nyt subscription with him. I ask him why he wants to pay for nyt, and dude sends me a wall of GPT generated text. To be fair he did admit it in a follow up, but it was just so gross to have to parse AI text in a personal chat. Nowhere is safe it seems. Sorry for the long rant
I'm sure that a lot of content produced is AI generated on Substack.
That being said I'm pretty skeptical of these checkers and their ability to keep pace. It's an arms race I am not sure can be won in the medium term.
People have been outsourcing their thinking and their words for a long time. AI is just a more efficient means of doing so.
Is it particularly sloppy and dishonest in the presentation of one's alleged writing? Of course. But just as AI sources from the "Public" that is reddit and the larger internet, people have been allowing the Public to think and speak for them forever, as Kierkegaard pointed out in detail.
All this is to say that most people were unoriginal parrots even before they had LLMs to do the parroting for them.
This is why I vowed early never to use AI in my writing. I even went so far as to write my second book of poetry offline, editing on a computer that does not go online. The text has never seen the internet.
When the first US Navy destroyers arrived in Queenstown, Ireland in 1917 to join the allied war against the Kaiser’s U-boats, the American commander reported his arrival to the British admiral at his headquarters. The admiral could not see him immediately because he was transcribing telegrams from his combat ships into longhand for his war report to the Admiralty in London. He was doing this because that is what Lord Admiral Nelson did when he was fighting Napoleon’s fleets. In 1805.
My mother worked on the B 52 program when it was being designed and the prototype was under construction. She received formulas and tables of numbers from the engineers and gave them back her calculations. Her job title was “Computer.” She computed.
That function is still in existence today, embedded in the various CAD/CAM systems engineers use to design modern aircraft.
Are today’s designs “less human” than they were in 1951 because the embedded computer is now digital?
When I was in high school and college we were prohibited from using electronic calculators in class. They would inhibit our ability to understand mathematics. That is what we were told.
When I was writing my Master’s Thesis (on heuristic AI, oddly) I was not allowed to submit a version written on a word processor. I had to type it on a typewriter or hire a professional secretary to generate the final, paper product. Typed on a typewriter. The physical act of writing was seen an important part of the academic process.
Nelson embedded the “Nelson touch” for decisive victory at sea into future admirals through longhand reporting instead of relying on telegraph slop.
My mother saved engineers from “computing slop.”
My high school teachers saved me from “mathematics slop.”
My thesis advisors saved me from “word processor slop.”
Dr. Richard Hamming, a Turing prize winner (and one of my professors) wrote “"Change does not mean progress, but progress requires change."
I offer that today’s slop is tomorrow’s “can you believe they actually did that?”
"When I was in high school and college we were prohibited from using electronic calculators in class. They would inhibit our ability to understand mathematics. That is what we were told."
Your teachers were correct.
Writing is different than mathematics as it relies on emotional rhetoric and appeals to common experience, something mathematics, as a purely abstract art, does not. AI tries to mimic it, but is fraudulent for the same reason having someone else write a letter to your wife and passing it as your own is fraudulent.
When I read academic writing from a century ago, I see references to Latin and Greek languages and texts. Untranslated, because everyone then knew those languages and the texts to which they referred. Toynbee is a recent example for me in my reading and research.
I lived with my maternal grandmother for a year in high school. She would quote Cicero and Tennyson at length from memory. That’s what she learned in school as a child.
Am I uneducated because I’m not literate in those? Is my writing inadequate because of that?
While that is an interesting tangential conversation, it doesnt address my point about AI writing having a lack of authenticity.
It does. What was authentic a century ago is not the same as authentic today. Today’s human writing would not be considered authentic a century ago.
I offer that what you define as authentic today will not be considered authentic in the future.
That is Hamming’s progress. You hit on it exactly with your opening example about craftsmanship. Today’s is not the same as yesterday’s. Where I differ from you is that I assert that today’s craftsmanship, and today’s AI writing are just fine; they have progressed. They are not the same as before but that that doesn’t make them “bad.”
Just like the format for Royal Navy reports, computers, and calculators. Different, yes. Bad, no. Better, yes.
Not having to hand tool metalwork because it is produced correctly is not bad. Not having to mash numbers to calculate the roots of a polynomial is not bad. Not having to labor for a week to create a 3000 word essay is not bad. All of those things represent technical, social, and cultural progress. They are all good.
The practitioners from the prior eras would probably not agree. “I had to learn Greek and Latin and memorize hundreds of lines of ancient prose; who are you to say you don’t have to?”
"It does. What was authentic a century ago is not the same as authentic today. Today’s human writing would not be considered authentic a century ago."
What do you mean by this? A love letter in the civil war era is just as authentic as a modern love letter. What has never been considered authentic is outsourcing the work to someone else.
"Claude, write my wife a love letter telling her I miss her" is far different than putting your emotions into words. If you can't see that I don't think I can help you.
We might be talking past each other.
I'm using an engineering history book written in 1898 for a project. Every chapter starts with a poem, a bible verse, a classic quotation, or other philosophical item relevant to the topic to be discussed in that chapter. And every chapter starts with a lengthy summary of the chapter (essentially a list of the headings). There are 40 chapters. It is a very technical history: pressures, temperatures, design drawings, operating instructions, etc. Not literature. Nothing like that is written today in that format or style. If I gave that author a modern engineering history, he would reject it as inadequate, incomplete, and stylistically unacceptable. Slop. If I input his text into a modern LLM, I would get a similar response, in the opposite vein. Most importantly, if I submitted that "authentic" 1898 format to a modern peer-reviewed journal, it would be rejected, regardless of the merit of my content. Slop.
I'll accept that love letters don't change much over the generations. The bulk of the writing I see is technical, programmatic, or policy-related for work, education, or advocacy. It is VERY attuned to the current day, practices, styles, and cultural mores. And those are constantly changing. LLM AI does a superb job of keeping up with those changes and that level of "progress," so I don't see it as slop. I see it as a valuable tool for meeting my daily performance requirements in my writing.
I'll concede the literature & personal letter slop points, mostly due to a lack of knowledge on my part. I do know that Sailors, soldiers, and prisoners have other people write their personal letters all the time. I hope that is not shocking to you.
A key point: LLM AI is not a person. It is a computer program. It is a tool, a machine, an inanimate object like a hammer, a leatherman, a calculator, or a spreadsheet. Having an LLM do something for me is using a hammer to knock off some rust, or a screwdriver to install a fastener. There is no moral or ethical conflict in the machine except in how I use it ( I can kill somebody with a hammer as easily as I drive a nail; the morality is mine, not the hammer’s).
For me, LLM AI text is not slop at all; like a calculator or a computer, it is an essential tool for getting my work done well, effectively, and efficiently. I do have to be able to use the tool; that can be a challenge for some people, especially at first, while they're learning it. As with any new tool.
I liked your post. Thank you for writing it. It surfaces an issue that accompanies all changes in technology, and has for all of human history. It’s good to get the topic on the table. I’ve read a thousand such posts & laments covering all the recorded millennia of human history and for all human technologies. Unfortunately, like your predecessors, I think you’re pushing a rope, so to speak. Dr. Hamming got it right. All progress requires change.
When I was taking our control systems class in graduate school, our grades were determined by the quality of the calculators we had purchased. The key activity in control systems is finding the roots of n-dimensional polynomials. Our ability to do that was 100% dependent on the quality (i.e. price) of our calculators. Our professor gave us brand and model numbers of the calculators needed to get a passing grade.
I'm not above naming one: NPOV checks as 100% AI on several articles. They're a Wikipedia watchdog, making their decision to employ AI rather involuted: We now have AI slop used as an investigative tool of AI slop. That's a lot of disrespect for Wikipedia. Maybe they're doing it right.
Heh, great minds think alike! I can’t stand any of it, in the classroom or in publications.
Funny story. I remember reading an obvious AI-counterfeit essay in Acton for whom I used to occasionally write, and I let the editor, Anthony Sacramone, know about it. I was nice about it and honestly thought he’d take my point in the proper spirit. Nope! He got pissed and told me that I couldn’t write for them anymore. I even apologized afterward for upsetting him (even though I was totally right and had a bunch of scans to prove it). But I’m still blackballed. So yeah, screw them.
I just wrote a piece in The Federalist about AI ruining education. All the kids cheat now and nothing is done about it. Worse, we make it easier by putting everything online and giving them college credit. Read more here: https://thefederalist.com/2026/06/01/artificial-intelligence-is-rewarding-high-school-grads-for-being-dishonest-frauds/
Acton Institute? I had several colleagues who worked there back in the day. Nice guys, but definitely some ego.
I don’t think it’s all AI slop at Acton, but it’s definitely happening with some of their lesser known “young” writers. And yes, I definitely encountered pretentiousness and anti-populist bias from the editor. This is pretty common with the self-serious boutique publications like Acton.
LOL! I took one at random and it was 60% AI!
https://blog.acton.org/archives/128443-europes-bureaucratization-of-parenting.html
They are like a time capsule to the 2000s at this point. They have not pivoted at all.
It sure is and this how we fight back:
https://substack.com/@michaelginsburg/note/c-269561061
Fight fire with (even more) FIRE!
I literally stopped using Substack for a few months because I was inundated with AI slop. I would love to get your thoughts on the project I’m starting! Take a look at my page to get a sense for what I’m thinking
I decided to test it with one of mine(I’m still skeptical and am on the free version) and it came back 100% human, so that is a mark in its favor so far.
Hi, I'm Dagmara. Thanks for the traffic.
Let me clear something up, because being filed next to crypto spam and get-rich-quick slop is a stretch, even for the internet.
I'm a PA-C with a PhD and more than twenty years in clinical medicine, ten of them in an emergency room. Every lab range on my Substack is mine. Every protocol is mine. Every opinion is mine, built from my own training and the years I've spent reading studies most people scroll right past.
Yes, I use AI. Here's the part the detector can't tell you. There's a difference between using AI and training it, and that difference is the whole ballgame.
Using AI is asking the machine what it thinks. It answers with the average of the internet, and the average of the internet is how you get slop. Confident, tidy, and often wrong about your health.
Training it is the opposite. I've fed it my standards, my ranges, the studies I trust, two decades of clinical reasoning, and now it checks my work against me instead of against Google. It flags when a number looks off so I catch it before 10,000 people read it. It tells me when a study doesn't say what I think it says. It cleans my grammar on the way out.
There's a name you already trust for that. Spellcheck. A spellchecker flags "teh" because someone loaded a dictionary into it, and nobody has ever called that cheating. I loaded twenty years of medicine into the tool instead of a dictionary. Same idea, more horsepower.
So no, the machine does not write for me. It helps me answer hundreds of comments from people asking for real help, because there is one of me and thousands of them, and I'd rather reach all of them well than a few of them slowly.
You brought up your grandfather carving birds by hand, and that you treasure them. I get that. I spent eleven years inside the cancer system supporting my daughter, and in that time I built a framework that now helps thousands of women understand their own bodies for the first time. Using a power tool to sand the edges doesn't make the bird less his. The hand and the mind are what make it real, and those are mine.
Here's the irony I keep turning over. You ran my Substack through a tool to decide I lean on tools. The difference is I read my work before I publish it. You didn't read mine before you judged it.
Dagmara
For those at home wondering, this comment was flagged as AI slop, but I doubt you needed me to tell you that.
Your response is LLM.
Sigh… did you even read what I wrote? I home school too. I was trying to actually connect but it doesn’t seem that’s your intention.
If connection is your goal why have your comment be written by an ai? Thats what it sounds like. If you did write it and want connection with human beings why do you write like an ai?
Be very careful about your accusations and beliefs. You may have a tool that you think is smart, and is giving you the truth, and that this truth is yours to crusade with, but Jordan Dann’s writing is all her own. Her thoughts, her research, her years of professional practice and development. How do I know? Because I watch her week after week painstakingly crafting what she shares with her audience for hours upon hours in her office. I bring her the coffee because she is so engrossed in her writing. I walk the dog so she can spend more time revising. We discuss her ideas on our couch. I see the process. Yes, AI is helpful in grammar and spelling and in other useful ways. But please do not trample on her work in some finger-pointing “Aha!” fashion. It is misguided.
My dilemma on utoob is that sometimes there is something I didn't know about, military history for example. I start the video. 30 seconds in it is blatantly obvious AI, yet the subject matter is something I want to know about, but I am angry at the content "creator" for using AI. Do I click away immediately, or stick around to learn something? Should I bother to complain about AI use in the comments?
I *really* do not want AI crap on Substack, that's not what I'm here for.
If it's any consolation, I'm convinced that the proliferation of AI is a fad. While it can be a useful tool and it will continue to be used in some areas, people hate it. I think we're on the left side of a Gaussian curve. Usage will increase, peak, and then fade. Grok is a money pit for X. The data center craze is hated and cannot be supported by our electrical infrastructure. For now we can complain, mock, avoid, and protest until the wave inevitably passes.
I get it - the AI fingerprints are really making Substack quite grimy. But I think your approach is wrong. You berate those who outsource their writing/thinking to machines but you are outsourcing your sense for quality, your taste, to Pangram. If the human connection is what you pine for, shouldn't you be able to judge whether that connection exists just by reading the posts? Why give a machine such sway over your judgement?
The AI checker tools can't tell whether a human writer is involved for the same reason that LLMs can't produce writing that feels human - because neither has any understanding of what it is to be human. You've probably outed some hard working authors who just write in an old-school, Strunk's-Elements-of-Style manner, which LLMs are expert at imitating, rather than in a more personal, idiosyncratic way, which LLMs can't do.
"If the human connection is what you pine for, shouldn't you be able to judge whether that connection exists just by reading the posts? Why give a machine such sway over your judgement?"
If one is on alert for AI writing, it often takes a decent amount of reading to discern AI writing. If you are on alert for AI writing, it is harder to enjoy the article. Either way is non-optimal, and the checker is the best way to ensure my time is not being wasted and I'm reading a human author.
"You've probably outed some hard working authors who just write in an old-school, Strunk's-Elements-of-Style manner, which LLMs are expert at imitating, rather than in a more personal, idiosyncratic way, which LLMs can't do."
Not how the modern tools work.
Great article, and very timely. I do think you should publicly expose the accounts in our sphere who are predominantly using AI, if only as a public safety measure. People need to know what they’re reading, I for one would love to know whether the stuff I read is actually written by a person or not.