Wow this exactly is what I noticed myself, especially when i picked up AI-isms. I was baffled when I started noticing that my own speech was mimicking what I heard on social media or when my writing started sounding robotic because I was reading so much produced by ai. I feel like I am still retraining myself to sound human again
When I was a dean I told my writing faculty in 2022 that their job would utterly change with ChatGPT. They thought they would become more important. I said sure but differently important. Their jobs would go back to being about grammar and structure and prosody, not "how do you feel in words." I mostly believed it when I said it; I wholly believe it now.
Totally agree with you. The devil's in the details and only by paying attention to them can you discern the nuts and bolts that are the remnants of machine-crafted writing
The real question isn’t whether technological constraints shape expression—they always have. It’s whether these particular constraints will prove generative over time, or merely restrictive. And we might not know the answer for another decade.
The color chart illustrates what tech is doing to the world, what AI is accelerating. The color of life is being sucked out of us to feed the algorithm and train AI. And when we let it, this draining of all color affects not only the world but our souls. The world is being remade in the image of the machine resulting in a barren landscape of Mordor. This is what we are pushing back against.
I am becoming more adept at recognizing AI writing. First thing I look for is your #1--the "it isn't just this, it's that" construction. That is a dead giveaway, when you see it more than once in any essay. And usually with AI it's in every third or fourth paragraph. But I'm also seeing the flatness, the lack of affect, the absence of any sign of soul, heart, or feeling at all. Well-ordered words, as if that is what "good writing" consists of. But once I get that flavor--or lack of flavor, more accurately--I am out of there no matter how interesting the topic is to me. I want to read what a human being says, how a human mind parses a subject, not how a robot spews forth sentences that sound like human speech.
Sounds quite accurate to my own experience. Though it was impressive at first I'm already finding Chat GPT to be predictable and shallow. It's becoming a bit like a Google that is easier to talk to and better at tracking things down (though be careful when it makes stuff up!).
Some of these patterns are not AI tells but rather part of what AI language models have learned from how certain sectors/fields speak and write.
My own use of dashes, e.g., i.e., "clearly/arguably" etc is longstanding and picked up, in part, from training in academic writing in my grad/PhD programs
I started writing as a new hobby in mid-life. I joined a writing group, began a daily journal practice, and listen to lectures by English professors whenever I have the chance.
What’s interesting/annoying is some of the long-standing basics that writing teachers pass along to novices are some of the same things people now accuse of being ChatGPT specific.
For example, the rule of three. Or using compare/contrast to persuade a reader. After putting forth a lot of effort in studying the complex sentence structure of classic novels, now we are being told that sentences containing a multiplicity of commas and clauses is a dead giveaway that ChatGPT was involved. As to the em dash of it all, just for funsies I cracked open my copy of Middlemarch and did a quick flip-through. Hardly a page exists without one, if not multiple, em dashes.
I despise ChatGPT—but I also am frustrated with the generalizations. 😆
Pretty much all of my pre-2020 writing in replete with the dash. I love it. My PhD adviser used it a ton too…which points to what I’m saying: AI language models are built on what people do, not vice versa
(I have Middemarch too and can confirm your observation)
Wow this exactly is what I noticed myself, especially when i picked up AI-isms. I was baffled when I started noticing that my own speech was mimicking what I heard on social media or when my writing started sounding robotic because I was reading so much produced by ai. I feel like I am still retraining myself to sound human again
It's really scary. Garbage in, garbage out is a rule of life, I fear.
Thanks for pointing us to Hollis Robbins’s excellent piece: https://substack.com/@hollisrobbins/p-170844031
That’s exactly the piece I was thinking about yesterday
She is one of one!
When I was a dean I told my writing faculty in 2022 that their job would utterly change with ChatGPT. They thought they would become more important. I said sure but differently important. Their jobs would go back to being about grammar and structure and prosody, not "how do you feel in words." I mostly believed it when I said it; I wholly believe it now.
Totally agree with you. The devil's in the details and only by paying attention to them can you discern the nuts and bolts that are the remnants of machine-crafted writing
The real question isn’t whether technological constraints shape expression—they always have. It’s whether these particular constraints will prove generative over time, or merely restrictive. And we might not know the answer for another decade.
amen
You might like my take https://thepuzzleanditspieces.substack.com/p/cold-text-gary-numan-and-the-synthesiser which has a very different framing
The color chart illustrates what tech is doing to the world, what AI is accelerating. The color of life is being sucked out of us to feed the algorithm and train AI. And when we let it, this draining of all color affects not only the world but our souls. The world is being remade in the image of the machine resulting in a barren landscape of Mordor. This is what we are pushing back against.
Fantastic title. Quite a useful list - I’d only ever heard the first one. I’ll have to add the rest to my mental roladex.
ChatGPTom is creepy!
Thank you so much. Reader beware!
I am becoming more adept at recognizing AI writing. First thing I look for is your #1--the "it isn't just this, it's that" construction. That is a dead giveaway, when you see it more than once in any essay. And usually with AI it's in every third or fourth paragraph. But I'm also seeing the flatness, the lack of affect, the absence of any sign of soul, heart, or feeling at all. Well-ordered words, as if that is what "good writing" consists of. But once I get that flavor--or lack of flavor, more accurately--I am out of there no matter how interesting the topic is to me. I want to read what a human being says, how a human mind parses a subject, not how a robot spews forth sentences that sound like human speech.
Sounds quite accurate to my own experience. Though it was impressive at first I'm already finding Chat GPT to be predictable and shallow. It's becoming a bit like a Google that is easier to talk to and better at tracking things down (though be careful when it makes stuff up!).
Admittedly, I have a penchant for puns, but I just love the headline.
Thank you!
Some of these patterns are not AI tells but rather part of what AI language models have learned from how certain sectors/fields speak and write.
My own use of dashes, e.g., i.e., "clearly/arguably" etc is longstanding and picked up, in part, from training in academic writing in my grad/PhD programs
Agree.
I started writing as a new hobby in mid-life. I joined a writing group, began a daily journal practice, and listen to lectures by English professors whenever I have the chance.
What’s interesting/annoying is some of the long-standing basics that writing teachers pass along to novices are some of the same things people now accuse of being ChatGPT specific.
For example, the rule of three. Or using compare/contrast to persuade a reader. After putting forth a lot of effort in studying the complex sentence structure of classic novels, now we are being told that sentences containing a multiplicity of commas and clauses is a dead giveaway that ChatGPT was involved. As to the em dash of it all, just for funsies I cracked open my copy of Middlemarch and did a quick flip-through. Hardly a page exists without one, if not multiple, em dashes.
I despise ChatGPT—but I also am frustrated with the generalizations. 😆
Pretty much all of my pre-2020 writing in replete with the dash. I love it. My PhD adviser used it a ton too…which points to what I’m saying: AI language models are built on what people do, not vice versa
(I have Middemarch too and can confirm your observation)