Mild correction: Pangram is AI. It’s not generative AI, but it is a large language model—a neural network trained to solve language problems. The difference is that it’s specialized for classification rather than being a general-purpose text generator.
I’m a “social influencer no one” who has been saying this for almost two years. The dishonesty and lack of transparency is SO at the heart of this. You wrote a book or an essay with AI help? Great. Tell your readers that before they fork over their money OR time for human words that are actually robot words the human sourced out because they’re in it for the money and not the connection. 😬
And Pangram? Anyone who doesn’t turn on that Pangram tool when I look for it gets blocked by me.
For two reasons:
1) If you’re someone doing it because it can “steal” from you? I don’t want to read any of your thoughts because they’re not very good ones if this is your take after the .0001% of reading you’ve done on the matter. (Our words are scraped by all the other bots coming at us from all over. Having an issue with Pangram is giving “Microchips are in the vaccines to track us!” while you don’t think one whit about that phone of yours you’re attached 24/7.)
2) If you’re someone turning it off and say so with fake and loud outrage over “witch hunts”—that is such pro-AI language!—you actually *are* the AI-using liar of a lying witch.
I don't undertand why people don't have Pangram turned on. I find it interesting to scan my work and have it read as written by a human. So far I haven't had it tell me otherwise. When I find certain pacing or tones in someones work on Substack, I want to use the Pangram tool to ensure that a human actually wrote it. Why not provide me with that reassurance as an author? Shouldn't you want to?
I like it scanning my work too! I write the way I speak: a little wonky and unpredictable like a hunchback hermit beauty contestant. 😋 If Pangram EVER comes back telling me I’m even 1% AI, that’s the day I tell Pangram their systems have expereinced catastrophic failure.
Thank you. I loved this piece, and I think the dishonesty angle cuts right to the heart of the issue. It's amazing how those covertly passing off AI work are nevertheless being very honest about something: they know how devalued AI output is, and so they conceal it. Otherwise, why not admit what they're up to?
Yep. Agree with all of this. The splits are about as you describe them. I think a lot of AI writers think “they can’t possibly know I used AI to edit my work, or to ‘ideate’ with”. I would like to ask these pearl clutching protesters “do you have a subscription to Claude or ChatGPT? On your computer right now would you have Claude book marked or the app installed?” I think these folks have lost sight in f where the boundary is. A bit like a cheating husband who thinks “she can’t possibly know” when actually we know what you’ve been doing this whole damn time. Stop with the lying.
I absolutely agree. AI is a tool. I don’t care if you use it, but just be honest about it. I can decide for myself when and how to extract value from an idea, regardless of how it was put into words. But don’t ask an LLM to produce a complete article, slap your name on it, and then cry foul when you’re flagged as less than 100% human. It seems simple to me.
Thank you for this. It made me look hard at my own writing which I admit is AI assisted … perhaps I’m just being intellectually dishonest. This gave me a lot of food for thought.
(And if you really don’t want to trip Pangram, you can always ask AI to flag editing problems and then you go fix them yourself. This is what my journalist friend told me their conclusion was.)
I really wish people would differentiate between "editing" and "revising". Because when I hear editing, I don't picturing going in and making extensive changes yourself, but pointing out to the author what might need changing so they can decide if and how. If you feed your work into LLM and accept it's output, you're no longer the sole writer, because revision is an important part of writing.
So, here's a blog post of mine from 2018 - the link is to the internet archive so it is easy to check when it was scraped and archived. Pangram rates it as 73% AI - feel free to scan it yourself. I am afraid that the tool is very much NOT foolproof, especially for anyone who received a formal education in writing or who is bilingual.
Thanks for running this through Pangram again. Obviously the new version is better, but as rece as a few weeks ago it was rating it as 73% as I stated. Unfortunately I can't attach the screenshot I took back then. I still think that automatically equating AI-assisted with bad is not sensible.
What I find most galling are the ones who claim they need AI because they are neurodivergent and then I look at the rest of their site and it's all utterly neurotypical, but yet another trendy label these bandwagon jumpers attach themselves to.
Hello, great stuff. A lot of succinct statements I can really empathise with.
I am fighting against the normalisation of AI reliance in scientific writing. My issue is both honesty (i.e. don’t get my sign off on something you didn’t write) and expectation (i.e. I don’t want my name on anything which isn’t 100% human). My fight isn’t going well, but I recently posted a plot outline for a direct to streaming movie
as part of managing my own annoyance (The Art of Writing and The Non-Art of In Silico Mimicry).
LOVE the intellectual dishonesty angle to the Pangram backlash. Spot on.
Fred Brooks wrote in his 1986 paper "No Silver Bullet", in a relatively brief section about AI:
"The hard thing about building software is deciding what to say, not saying it."
Brooks was awarded a Turing in 1999.
Mild correction: Pangram is AI. It’s not generative AI, but it is a large language model—a neural network trained to solve language problems. The difference is that it’s specialized for classification rather than being a general-purpose text generator.
I don't know about that. LLMs record, retain, and generate language. Pangram does not, and they expressly say their product is not a LLM.
I’m a “social influencer no one” who has been saying this for almost two years. The dishonesty and lack of transparency is SO at the heart of this. You wrote a book or an essay with AI help? Great. Tell your readers that before they fork over their money OR time for human words that are actually robot words the human sourced out because they’re in it for the money and not the connection. 😬
And Pangram? Anyone who doesn’t turn on that Pangram tool when I look for it gets blocked by me.
For two reasons:
1) If you’re someone doing it because it can “steal” from you? I don’t want to read any of your thoughts because they’re not very good ones if this is your take after the .0001% of reading you’ve done on the matter. (Our words are scraped by all the other bots coming at us from all over. Having an issue with Pangram is giving “Microchips are in the vaccines to track us!” while you don’t think one whit about that phone of yours you’re attached 24/7.)
2) If you’re someone turning it off and say so with fake and loud outrage over “witch hunts”—that is such pro-AI language!—you actually *are* the AI-using liar of a lying witch.
Anywho.
I hope that your essay finds traction!!
I don't undertand why people don't have Pangram turned on. I find it interesting to scan my work and have it read as written by a human. So far I haven't had it tell me otherwise. When I find certain pacing or tones in someones work on Substack, I want to use the Pangram tool to ensure that a human actually wrote it. Why not provide me with that reassurance as an author? Shouldn't you want to?
I like it scanning my work too! I write the way I speak: a little wonky and unpredictable like a hunchback hermit beauty contestant. 😋 If Pangram EVER comes back telling me I’m even 1% AI, that’s the day I tell Pangram their systems have expereinced catastrophic failure.
Interesting that Pangram is willing to pay a bounty. Thanks for linking to my essay!
Thank you. I loved this piece, and I think the dishonesty angle cuts right to the heart of the issue. It's amazing how those covertly passing off AI work are nevertheless being very honest about something: they know how devalued AI output is, and so they conceal it. Otherwise, why not admit what they're up to?
Yep. Agree with all of this. The splits are about as you describe them. I think a lot of AI writers think “they can’t possibly know I used AI to edit my work, or to ‘ideate’ with”. I would like to ask these pearl clutching protesters “do you have a subscription to Claude or ChatGPT? On your computer right now would you have Claude book marked or the app installed?” I think these folks have lost sight in f where the boundary is. A bit like a cheating husband who thinks “she can’t possibly know” when actually we know what you’ve been doing this whole damn time. Stop with the lying.
I absolutely agree. AI is a tool. I don’t care if you use it, but just be honest about it. I can decide for myself when and how to extract value from an idea, regardless of how it was put into words. But don’t ask an LLM to produce a complete article, slap your name on it, and then cry foul when you’re flagged as less than 100% human. It seems simple to me.
Thank you for this. It made me look hard at my own writing which I admit is AI assisted … perhaps I’m just being intellectually dishonest. This gave me a lot of food for thought.
Yes 💯💪🏼
(And if you really don’t want to trip Pangram, you can always ask AI to flag editing problems and then you go fix them yourself. This is what my journalist friend told me their conclusion was.)
I really wish people would differentiate between "editing" and "revising". Because when I hear editing, I don't picturing going in and making extensive changes yourself, but pointing out to the author what might need changing so they can decide if and how. If you feed your work into LLM and accept it's output, you're no longer the sole writer, because revision is an important part of writing.
Great citations -- that Bona report is really interesting and it should be on more people's radars. Thanks for sharing it.
So, here's a blog post of mine from 2018 - the link is to the internet archive so it is easy to check when it was scraped and archived. Pangram rates it as 73% AI - feel free to scan it yourself. I am afraid that the tool is very much NOT foolproof, especially for anyone who received a formal education in writing or who is bilingual.
https://web.archive.org/web/20240614192939/http://jimako.com/posts/2018-08-12-aus-encryption/
Pangram gives me 100% human on this post. https://www.pangram.com/history/718b83b9-fc74-43b2-b461-c5677adac85d?ucc=aLPfG69yzyI
Thanks for running this through Pangram again. Obviously the new version is better, but as rece as a few weeks ago it was rating it as 73% as I stated. Unfortunately I can't attach the screenshot I took back then. I still think that automatically equating AI-assisted with bad is not sensible.
What I find most galling are the ones who claim they need AI because they are neurodivergent and then I look at the rest of their site and it's all utterly neurotypical, but yet another trendy label these bandwagon jumpers attach themselves to.
Hello, great stuff. A lot of succinct statements I can really empathise with.
I am fighting against the normalisation of AI reliance in scientific writing. My issue is both honesty (i.e. don’t get my sign off on something you didn’t write) and expectation (i.e. I don’t want my name on anything which isn’t 100% human). My fight isn’t going well, but I recently posted a plot outline for a direct to streaming movie
as part of managing my own annoyance (The Art of Writing and The Non-Art of In Silico Mimicry).
Wow. Honestly this was a journey through several angles I had not explored and appreciate the examples.
And, I guess I’m also disappointed that what I had read from trusted writers on their flags was maybe not 100% above board.
People used to add the AI note, just like there used to be that default apology for emails typed from a phone.
I guess after awhile, we lost track of that line and meaning and AI detection stirred the pot.