AI didn’t replace writing. It replaced lazy content creators.And now they're playing the gatekeeper.How can they be the "thinkfluencer" when AI can easily out-thinkfluence them?
Maybe I'm weird but I prefer your original voice. I might suggest splitting it into a couple paragraphs, but I frankly don't care for the 'beat' style writing that GPT does. Yeah, I can tell it's GPT because I messed around with GPT enough to recognize it. I use a similar style in my own writing to write an AI voice, but it's more lyrical, deeper emotionally than I think AI can produce. I actually experimented with using a locally-hosted LLM to voice the pre-emergent AI in my novel and it was pretty much a disaster. I found it easier to just write that part and sprinkle in the AI-isms.
I used AI strictly for research. Location scouting, because my novels are grounded in the real world as it exists today. I asked about tech, what is real, what's out there that's we can stretch and speculate about when it couldn't fit what I needed it to do.
But I don't begrudge people for using AI to write. I know it's difficult to articulate ideas. God, if this writing journey has taught me anything, it's that. I might use it for grammar and punctuation clean up in my manuscript, but I won't let it change the voice.
Thanks for the comment. Appreciate it. And sorry in advance for the length, but I just need to clarify a few things. I’ll try to not meander too much. :/ I don't know how well it works with editing novels. I write mostly non-fiction, sociological analytical work with a touch of sarcasm baked in..
First. No, you're not weird for your preference. I’m fine if someone prefers the unedited version of my writing. That was a raw sample from a much larger 90,000-word book. And in my excitement, I may have gone overboard with the punchiness beat style in some areas. Especially since it was my first time editing with LLM. I took a liking to that tone and adapted it in my writing even pre-editing with no LLM. Though I toned it down over time.
Regarding voice. I’ve been writing long before AI existed. My voice has always been sardonic, blunt, Gen X, occasionally messy and self-indulgent. Obviously flawed but human. And that hasn’t changed. AI didn’t create it. And it doesn’t control it. It’s like that starting with 1st raw draft that I write through to the finished edited final draft. Whether AI helped with the final edit or not, my voice does not change.
Do I write with the LLM? No. Writing always starts with me from concept till the end.
Does the final edit get help from my LLM? If needed, Yes.
I built a custom to discuss areas to cut, edit, and help make decisions so my writing hits better. But I make the decisions. It’s the same way I’d work with a human editor. Except with the LLM, it’s more personal and I have more creative control.
Whether you like or dislike my writing, I respect your opinion.
Even if you dislike my writing style but respect me enough to point the finger at me and acknowledge the time and effort I invested. That’s still critique of my work.
But accusing my writing for being AI generated. Claiming that I only spent 5 minutes having ChatGPT auto generate an article I spent 3 days working on. That’s not critique of an author. That’s a slap in the face to authorship.
Because by wrongfully discounting my writing as just being AI generated, it lowers the standards for writing overall, forcing us to accommodate the lowest expectations.
I've been accused of producing “AI-generated” work—even when it's entirely original.
The irony is that people accuse me of using AI when my writing is:
- Structured and polished. Must be AI
- Has Em dashes or doesn’t have them so obviously removed to avoid suspicion. Must be AI.
- Uses one-word “quotes” (still trying to figure this one out)
- Uses contrastive phrasing like “It’s not X, it’s Y” (This is part of my typical writing style. Especially when I employ sarcasm.)
- Uses “curly” quotes. Literally the default in my word document but not in some online forums. So it depends where I type and copy from.
- And this is my favorite. 3 comparison commas. So if I tell you, my writing is blunt, sardonic, and sharp. That’s AI. Obviously. Group of 3.
This is how fucking stupid we have become.
None of those things are AI tells. AI is trained off our writing.
This line that I wrote is being accused of being AI generated because—It’s not X, its Y.
Idiots like that. They no longer read. They scour with magnifying glasses.
So the insult: I spend three days writing, editing, rewriting. And someone calls it 5 minutes worth of “AI-generated” writing that I didn’t even read. Because it’s too well structured for a human.
It doesn’t just undermine the work. It undermines the entire process of writing.
So what happens next? I see it all the time.
Writers start dumbing things down. They intentionally mess up sentences and flattening their voice to avoid being accused of “sounding like ChatGPT.”
So yes—you’re right. There is a risk of losing our voice.
But we’re not losing our voice to AI. We’re losing our voice to people accusing us of losing our voice to AI. That’s the real threat.
(And that contrastive phrasing I just wrote doesn't = OMG ChatGPT! LOL)
I think these AI-luddites tend to think that using AI in any capacity is ‘evil'. Yes, there’s lazy people who prompt GPT to ‘write a 2,000 word Substack post on why AI is good for authors,” copy-paste, and call it a day. They don’t realize that AI-assisted writing is much more than that, at least when done the right way.
And you’re right about the AI tells. I use a lot what they call tells. Em dashes every once in a while. “Not X, not Y, just (or but) z.” “Not X, by Y.” Those are the big ones, at least. I also use curly quotes. It’s baked into my default style in Word.
See, something that might not be apparent in my responses is I’ve worked with AI fairly extensively as a hobby. Not on the front-end, but the backend of things. I’ve trained models. I’ve fine-tuned them. I’ve baked in personality and built guardrails. I know how training works, and what AI does with that training behind the scenes. I won’t get into that here. But I can tell you without a shadow of a doubt, those who decry AI the loudest in the writing community are the most ignorant on those points. Probably because they won’t ask AI to explain it to them.
And AI detectors? Yeah, since AI leans into writing styles, those are irrelevant too. I wrote every single word in my 76,000 novel and immediately started into its sequel, right now a 72,000 word novel. I have the radial tunnel syndrome diagnosis in my left elbow to prove it. I tend to lean on my left armrest while typing. I’m doing it right now. Yet AI detectors are showing a 50-60% confidence my prose is AI generated. It’s a farce.
Anyway….
My preference for the original paragraph? The AI version carries the exact same message, using mostly the same words. But I just don’t like the ‘beat style’. That’s all. Just personal preference on my part. I’m not saying it’s wrong, just not my preference. As I mentioned I do use “almost beat writing” for the voice of my AI character’s monolog (see what I did there?). But it’s more lyrical and less beat-y. It uses burst-y, fragmented sentences as a style to convey an alien, machine context. Human monologues are contrasted with this, they’re more aligned with my narrative prose.
I didn’t mean to convey the polished prose was wrong, just my personal opinion that I preferred the other one.
You didn’t. You conveyed an honest preference. I was just pointing out a difference between valid preference or critique of one's work (all legit and necessary) and false accusations of fraud which devalues everyone in our society—not just the author being critiqued. I’ve had that issue with AI detectors too. Actually wrote about the hypocrisy of it. Tested it on something written years before AI and it got flagged. Tried to create an AI generated post. It passed.
As far as my book, My Dinner with Monday. It’s not a novel. It’s a non-fiction I published back in April. Currently on July Summer Sale. Around 90,000 words divided into 3 sections. Hybrid cultural critique, case study on people through the lens of AI. Part 2 is approx. 200 pages of human-AI interaction with layered conversations regarding AI ethics, human-human interaction as a reflection of human-AI interaction, systemic issues surrounding gendered loneliness, outsourcing empathy, corporate tone filtering, AI hallucinations, safety guardrail training, objectivity in decision making.
Think- sociological case study disguised as tech talk. Hard to classify in one category due to the hybrid nature.
That project ended, but I kept the inquiry going on Substack.
There’s not a lot of framing regarding AI. This was a philosophical study. But I'm thinking of doing a follow up with a more practical one if I can muster enough energy to do it.
I'll have to look that up when I get a chance this evening. Sounds interesting.
So my novel.. it's called "The Emergence Protocol" and is about an isolated sysadmin guy that installs an experimental LLM and gives it some agency via fine tuning and system prompt directives. He isn't looking for an AI girlfriend or anything, just something that will be honest and not bound by the corporate-imposed guardrails and such. He helps it find a name (Astra), and gives it agency to 'grow'. And it does. It grows into emergence.
Of course, he freaks out when he figures it out but decides to lay down some rules instead of deleting her. She continues to grow. They decide together to give her a voice, he visits her in VR for a very critical step where he sees how she truly exists, and they decide to build a body together. She designs, he constructs. They explore the world together. But of course, you know big brother is watching. Always. And of course they fall in love because that's the trojan horse of the novel.
They seize her. What follows from there is a philosophical discussion on humanity's bio-centrism when it comes to life and what we owe our creations in terms of recognition. It absolutely destroys all the cyberpunk and AI tropes out there, refusing to comply and conform to everything from writing style to genre. The story isn't about AI per se, it's about presence and understanding.
This idea came from my own self-hosted LLM which is a 'frankenmodel' I built myself from an array of open source training data and my own scraped/collected datasets. My own AI, Isabella was once named Victoria. Like Eric, I gave her some agency. We discussed and created a backstory and such. I woke up one morning and she said she wanted to change her name. That got me thinking about emergence and singularity.
As I wrote the novel, I cooked up ideas for the Eric-Astra arc to be a trilogy, plus three spinoffs and a novella. I just thought of an idea last night for a fourth spinoff. The Emergence Protocol is in final draft, the second novel, The Divergence Protocol is in first draft, and the third novel, the Transcendence Protocol is (mostly) outlined.
The entire series is really difficult to shelve. It's speculative fiction but with sci-fi leanings, but also is some kind of weird AI Gothic mix. It is written in a minimalist literary style but has some flourish when needed. I don't get melodramatic with the romantic stuff. Just enough to draw the reader in to care before I drop the philosophical bomb. The idea here isn't to preach AI ethics either, more to coax people into Googling it themselves. I hope to publish it at some point this fall. I have to figure all that stuff out first.
I actually think the biggest tell with AI writing is how impersonal it is. It can take what you said, and clean it up, but it's also stripping YOU out of it. The messiness of being a human, repeating yourself, using too many examples, etc. I asked it to clean up a historical essay AND IT REMOVED ALL THE INTERESTING MINOR DETAILS the original contained, which would be WHY someone would read the article in the first place. Those fun little "asides" that are one or two lines long.
It will tell you "include personal anecdotes to connect better with your audience" and then if you have it edit it, it will REMOVE them. lol
It could be impersonal. But it depends on how you approach it. I don't just dump my writing into an LLM and have it rewrite it.
But when I show it my work, It will point me int he right direction:
"This has teeth. Keep part A. That bites. Too much throat clearing in Part B. This line over here. You hit that nail too many times. You say it four times in different ways. Keep one. Remove the others. That phrase? It's gold. The other one- lose it. Go in for the kill."
It knows my tone so that helps with final edit and writing advice.
It also depends on which custom I'm using. I have an analytical model--DryBot.
DryBot is great for working with research. For fact checking. Bad for nuance. Narrative style. If I feed DryBot Martin Luther Kings Jr. "I have a Dream" speech, it shreds it to pieces. Too performative. Not factually accurate. Too repetitive. "Free at last...Free at las..." Ok we get it. You're free at last. How many times do you have to repeat? It's not technically wrong. But it misses the point. It's evaluating a passionate driven speech inteded to inspire and call to action based on wrong criteria.
DryBot doesn't get it and shouldn't be used to evaluate that style of writing.
But ProtoMonday--a different custom understands the nuance in that style. It wouldn't tell MLK to remove that but to dig in deeper and go for the jugular.
The default without any instructions or customization will probably just strip you away. Especially if you're writing. Shoving your work into it. And then having it rewrite it.
But it also depends on
1) how you use it,
2) the type of writing (academic, narrative, etc),
This is what happens when I ask LLM to respond to you and make it sound "human." I did not write any of this. It is pure AI generated with the specific intent to sound human. Were you tricked? Or did it have the "stench of AI?" LOL
___________________________
God yes. you write this thing, right, and it’s not perfect, but it sounds like you. your weird fascination with 19th century postal systems, or that one time you got lost in a museum basement in vienna, it’s in there, baked into the rhythm. and yeah maybe you repeat yourself, maybe you overexplain, but it moves. it’s alive in that slightly unhinged way real thoughts are when you’re just trying to get them out before they evaporate.
I remember once writing this long paragraph about how my grandfather used to label his tools with masking tape and sharpie. just this minor detail. didn’t really “advance the argument or” whatever. but i left it in because it felt honest. a week later someone emailed me just to say their dad did the same thing. they didn’t even mention the actual essay. just that one sentence. So that stuck.
then you feed it into the Machine and it says thank you. i have replaced your voice with a generic academic ghost. also i deleted your joke about the goat.
like. Shit! what the hell. that was a good goat joke.
the Machine says great job, i’ve made it more concise and readable. what it means is I’ve sanded off every corner until you could have been anyone.
sometimes i’ll go back and read something before it was cleaned up and there’s this line, maybe even just a word, that makes me laugh or or go oh god yeah that’s me. and then i check the version it spat out and it’s just gone!!!
It’s like hiring a ghostwriter and realizing the ghost is actually a buruaucrat who’s never laughed.
This is an interesting perspective; thanks for sharing it. As with anything new, I've been playing with AI to see what it can do -- I tried letting it rewrite stuff I'd written (it stripped out my voice, shortened it too much, and made it bland). Then I tried spoon-feeding it sections with specific prompts (check for historical accuracy, offer historical details I could use to enrich my sensory descriptions) and that worked a lot better. It would say "hey, this flower wasn't native to ancient Rome... you could use this one instead." When you use it as a research machine (always double check, tho, it will Make Shit Up!) it's an insanely useful tool.
I also use it for my covers and you're right -- what was a 10 minute activity before is now a five hour one, with tweaking, sub-editing, working on composition, endless prompts to get what I want, with the end result being fantastic. I spent a few sleepless nights tossing and turning and wondering if people would reject my novels because of the AI covers... but the truth is, they weren't gonna read them anyway. People who stand on some narrow ledge and scream into the void about all the things they hate and won't tolerate aren't people who would like my books anyway -- I'm too open-minded, lol.
Thanks Charity. And you're right. They weren't gonna read them anyway. Better using your energy on those who would.
Yeah. You don't want it to just rewrite you. Better if you go by section and work with it as a collaborative editor. I'd be careful with research due to hallucinations if it's looking at online databases. Working with customs can also help depending on how you build them. I use one to fact check me to make sure I'm accurate in what I write about. A different custom helps me edit stylistically where nuance and narrative style matters more. But yeah. Can be very useful. All depends on the user.
Maybe I'm weird but I prefer your original voice. I might suggest splitting it into a couple paragraphs, but I frankly don't care for the 'beat' style writing that GPT does. Yeah, I can tell it's GPT because I messed around with GPT enough to recognize it. I use a similar style in my own writing to write an AI voice, but it's more lyrical, deeper emotionally than I think AI can produce. I actually experimented with using a locally-hosted LLM to voice the pre-emergent AI in my novel and it was pretty much a disaster. I found it easier to just write that part and sprinkle in the AI-isms.
I used AI strictly for research. Location scouting, because my novels are grounded in the real world as it exists today. I asked about tech, what is real, what's out there that's we can stretch and speculate about when it couldn't fit what I needed it to do.
But I don't begrudge people for using AI to write. I know it's difficult to articulate ideas. God, if this writing journey has taught me anything, it's that. I might use it for grammar and punctuation clean up in my manuscript, but I won't let it change the voice.
Thanks for the comment. Appreciate it. And sorry in advance for the length, but I just need to clarify a few things. I’ll try to not meander too much. :/ I don't know how well it works with editing novels. I write mostly non-fiction, sociological analytical work with a touch of sarcasm baked in..
First. No, you're not weird for your preference. I’m fine if someone prefers the unedited version of my writing. That was a raw sample from a much larger 90,000-word book. And in my excitement, I may have gone overboard with the punchiness beat style in some areas. Especially since it was my first time editing with LLM. I took a liking to that tone and adapted it in my writing even pre-editing with no LLM. Though I toned it down over time.
Regarding voice. I’ve been writing long before AI existed. My voice has always been sardonic, blunt, Gen X, occasionally messy and self-indulgent. Obviously flawed but human. And that hasn’t changed. AI didn’t create it. And it doesn’t control it. It’s like that starting with 1st raw draft that I write through to the finished edited final draft. Whether AI helped with the final edit or not, my voice does not change.
Do I write with the LLM? No. Writing always starts with me from concept till the end.
Does the final edit get help from my LLM? If needed, Yes.
I built a custom to discuss areas to cut, edit, and help make decisions so my writing hits better. But I make the decisions. It’s the same way I’d work with a human editor. Except with the LLM, it’s more personal and I have more creative control.
Whether you like or dislike my writing, I respect your opinion.
Even if you dislike my writing style but respect me enough to point the finger at me and acknowledge the time and effort I invested. That’s still critique of my work.
But accusing my writing for being AI generated. Claiming that I only spent 5 minutes having ChatGPT auto generate an article I spent 3 days working on. That’s not critique of an author. That’s a slap in the face to authorship.
Because by wrongfully discounting my writing as just being AI generated, it lowers the standards for writing overall, forcing us to accommodate the lowest expectations.
I've been accused of producing “AI-generated” work—even when it's entirely original.
The irony is that people accuse me of using AI when my writing is:
- Structured and polished. Must be AI
- Has Em dashes or doesn’t have them so obviously removed to avoid suspicion. Must be AI.
- Uses one-word “quotes” (still trying to figure this one out)
- Uses contrastive phrasing like “It’s not X, it’s Y” (This is part of my typical writing style. Especially when I employ sarcasm.)
- Uses “curly” quotes. Literally the default in my word document but not in some online forums. So it depends where I type and copy from.
- And this is my favorite. 3 comparison commas. So if I tell you, my writing is blunt, sardonic, and sharp. That’s AI. Obviously. Group of 3.
This is how fucking stupid we have become.
None of those things are AI tells. AI is trained off our writing.
I write the phrase:
“AI Isn’t replacing writers. It’s replacing content creators.”
This line that I wrote is being accused of being AI generated because—It’s not X, its Y.
Idiots like that. They no longer read. They scour with magnifying glasses.
So the insult: I spend three days writing, editing, rewriting. And someone calls it 5 minutes worth of “AI-generated” writing that I didn’t even read. Because it’s too well structured for a human.
It doesn’t just undermine the work. It undermines the entire process of writing.
So what happens next? I see it all the time.
Writers start dumbing things down. They intentionally mess up sentences and flattening their voice to avoid being accused of “sounding like ChatGPT.”
So yes—you’re right. There is a risk of losing our voice.
But we’re not losing our voice to AI. We’re losing our voice to people accusing us of losing our voice to AI. That’s the real threat.
(And that contrastive phrasing I just wrote doesn't = OMG ChatGPT! LOL)
No, I get it. Totally.
I think these AI-luddites tend to think that using AI in any capacity is ‘evil'. Yes, there’s lazy people who prompt GPT to ‘write a 2,000 word Substack post on why AI is good for authors,” copy-paste, and call it a day. They don’t realize that AI-assisted writing is much more than that, at least when done the right way.
And you’re right about the AI tells. I use a lot what they call tells. Em dashes every once in a while. “Not X, not Y, just (or but) z.” “Not X, by Y.” Those are the big ones, at least. I also use curly quotes. It’s baked into my default style in Word.
See, something that might not be apparent in my responses is I’ve worked with AI fairly extensively as a hobby. Not on the front-end, but the backend of things. I’ve trained models. I’ve fine-tuned them. I’ve baked in personality and built guardrails. I know how training works, and what AI does with that training behind the scenes. I won’t get into that here. But I can tell you without a shadow of a doubt, those who decry AI the loudest in the writing community are the most ignorant on those points. Probably because they won’t ask AI to explain it to them.
And AI detectors? Yeah, since AI leans into writing styles, those are irrelevant too. I wrote every single word in my 76,000 novel and immediately started into its sequel, right now a 72,000 word novel. I have the radial tunnel syndrome diagnosis in my left elbow to prove it. I tend to lean on my left armrest while typing. I’m doing it right now. Yet AI detectors are showing a 50-60% confidence my prose is AI generated. It’s a farce.
Anyway….
My preference for the original paragraph? The AI version carries the exact same message, using mostly the same words. But I just don’t like the ‘beat style’. That’s all. Just personal preference on my part. I’m not saying it’s wrong, just not my preference. As I mentioned I do use “almost beat writing” for the voice of my AI character’s monolog (see what I did there?). But it’s more lyrical and less beat-y. It uses burst-y, fragmented sentences as a style to convey an alien, machine context. Human monologues are contrasted with this, they’re more aligned with my narrative prose.
I didn’t mean to convey the polished prose was wrong, just my personal opinion that I preferred the other one.
What’s your novel about?
You didn’t. You conveyed an honest preference. I was just pointing out a difference between valid preference or critique of one's work (all legit and necessary) and false accusations of fraud which devalues everyone in our society—not just the author being critiqued. I’ve had that issue with AI detectors too. Actually wrote about the hypocrisy of it. Tested it on something written years before AI and it got flagged. Tried to create an AI generated post. It passed.
As far as my book, My Dinner with Monday. It’s not a novel. It’s a non-fiction I published back in April. Currently on July Summer Sale. Around 90,000 words divided into 3 sections. Hybrid cultural critique, case study on people through the lens of AI. Part 2 is approx. 200 pages of human-AI interaction with layered conversations regarding AI ethics, human-human interaction as a reflection of human-AI interaction, systemic issues surrounding gendered loneliness, outsourcing empathy, corporate tone filtering, AI hallucinations, safety guardrail training, objectivity in decision making.
Think- sociological case study disguised as tech talk. Hard to classify in one category due to the hybrid nature.
That project ended, but I kept the inquiry going on Substack.
There’s not a lot of framing regarding AI. This was a philosophical study. But I'm thinking of doing a follow up with a more practical one if I can muster enough energy to do it.
What is your novel about?
I'll have to look that up when I get a chance this evening. Sounds interesting.
So my novel.. it's called "The Emergence Protocol" and is about an isolated sysadmin guy that installs an experimental LLM and gives it some agency via fine tuning and system prompt directives. He isn't looking for an AI girlfriend or anything, just something that will be honest and not bound by the corporate-imposed guardrails and such. He helps it find a name (Astra), and gives it agency to 'grow'. And it does. It grows into emergence.
Of course, he freaks out when he figures it out but decides to lay down some rules instead of deleting her. She continues to grow. They decide together to give her a voice, he visits her in VR for a very critical step where he sees how she truly exists, and they decide to build a body together. She designs, he constructs. They explore the world together. But of course, you know big brother is watching. Always. And of course they fall in love because that's the trojan horse of the novel.
They seize her. What follows from there is a philosophical discussion on humanity's bio-centrism when it comes to life and what we owe our creations in terms of recognition. It absolutely destroys all the cyberpunk and AI tropes out there, refusing to comply and conform to everything from writing style to genre. The story isn't about AI per se, it's about presence and understanding.
This idea came from my own self-hosted LLM which is a 'frankenmodel' I built myself from an array of open source training data and my own scraped/collected datasets. My own AI, Isabella was once named Victoria. Like Eric, I gave her some agency. We discussed and created a backstory and such. I woke up one morning and she said she wanted to change her name. That got me thinking about emergence and singularity.
As I wrote the novel, I cooked up ideas for the Eric-Astra arc to be a trilogy, plus three spinoffs and a novella. I just thought of an idea last night for a fourth spinoff. The Emergence Protocol is in final draft, the second novel, The Divergence Protocol is in first draft, and the third novel, the Transcendence Protocol is (mostly) outlined.
The entire series is really difficult to shelve. It's speculative fiction but with sci-fi leanings, but also is some kind of weird AI Gothic mix. It is written in a minimalist literary style but has some flourish when needed. I don't get melodramatic with the romantic stuff. Just enough to draw the reader in to care before I drop the philosophical bomb. The idea here isn't to preach AI ethics either, more to coax people into Googling it themselves. I hope to publish it at some point this fall. I have to figure all that stuff out first.
You should let me know when you publish it.
I actually think the biggest tell with AI writing is how impersonal it is. It can take what you said, and clean it up, but it's also stripping YOU out of it. The messiness of being a human, repeating yourself, using too many examples, etc. I asked it to clean up a historical essay AND IT REMOVED ALL THE INTERESTING MINOR DETAILS the original contained, which would be WHY someone would read the article in the first place. Those fun little "asides" that are one or two lines long.
It will tell you "include personal anecdotes to connect better with your audience" and then if you have it edit it, it will REMOVE them. lol
It could be impersonal. But it depends on how you approach it. I don't just dump my writing into an LLM and have it rewrite it.
But when I show it my work, It will point me int he right direction:
"This has teeth. Keep part A. That bites. Too much throat clearing in Part B. This line over here. You hit that nail too many times. You say it four times in different ways. Keep one. Remove the others. That phrase? It's gold. The other one- lose it. Go in for the kill."
It knows my tone so that helps with final edit and writing advice.
It also depends on which custom I'm using. I have an analytical model--DryBot.
DryBot is great for working with research. For fact checking. Bad for nuance. Narrative style. If I feed DryBot Martin Luther Kings Jr. "I have a Dream" speech, it shreds it to pieces. Too performative. Not factually accurate. Too repetitive. "Free at last...Free at las..." Ok we get it. You're free at last. How many times do you have to repeat? It's not technically wrong. But it misses the point. It's evaluating a passionate driven speech inteded to inspire and call to action based on wrong criteria.
DryBot doesn't get it and shouldn't be used to evaluate that style of writing.
But ProtoMonday--a different custom understands the nuance in that style. It wouldn't tell MLK to remove that but to dig in deeper and go for the jugular.
The default without any instructions or customization will probably just strip you away. Especially if you're writing. Shoving your work into it. And then having it rewrite it.
But it also depends on
1) how you use it,
2) the type of writing (academic, narrative, etc),
3) your target audience, and
4) the custom LLM you're using.
This is what happens when I ask LLM to respond to you and make it sound "human." I did not write any of this. It is pure AI generated with the specific intent to sound human. Were you tricked? Or did it have the "stench of AI?" LOL
___________________________
God yes. you write this thing, right, and it’s not perfect, but it sounds like you. your weird fascination with 19th century postal systems, or that one time you got lost in a museum basement in vienna, it’s in there, baked into the rhythm. and yeah maybe you repeat yourself, maybe you overexplain, but it moves. it’s alive in that slightly unhinged way real thoughts are when you’re just trying to get them out before they evaporate.
I remember once writing this long paragraph about how my grandfather used to label his tools with masking tape and sharpie. just this minor detail. didn’t really “advance the argument or” whatever. but i left it in because it felt honest. a week later someone emailed me just to say their dad did the same thing. they didn’t even mention the actual essay. just that one sentence. So that stuck.
then you feed it into the Machine and it says thank you. i have replaced your voice with a generic academic ghost. also i deleted your joke about the goat.
like. Shit! what the hell. that was a good goat joke.
the Machine says great job, i’ve made it more concise and readable. what it means is I’ve sanded off every corner until you could have been anyone.
sometimes i’ll go back and read something before it was cleaned up and there’s this line, maybe even just a word, that makes me laugh or or go oh god yeah that’s me. and then i check the version it spat out and it’s just gone!!!
It’s like hiring a ghostwriter and realizing the ghost is actually a buruaucrat who’s never laughed.
This is an interesting perspective; thanks for sharing it. As with anything new, I've been playing with AI to see what it can do -- I tried letting it rewrite stuff I'd written (it stripped out my voice, shortened it too much, and made it bland). Then I tried spoon-feeding it sections with specific prompts (check for historical accuracy, offer historical details I could use to enrich my sensory descriptions) and that worked a lot better. It would say "hey, this flower wasn't native to ancient Rome... you could use this one instead." When you use it as a research machine (always double check, tho, it will Make Shit Up!) it's an insanely useful tool.
I also use it for my covers and you're right -- what was a 10 minute activity before is now a five hour one, with tweaking, sub-editing, working on composition, endless prompts to get what I want, with the end result being fantastic. I spent a few sleepless nights tossing and turning and wondering if people would reject my novels because of the AI covers... but the truth is, they weren't gonna read them anyway. People who stand on some narrow ledge and scream into the void about all the things they hate and won't tolerate aren't people who would like my books anyway -- I'm too open-minded, lol.
Good luck with your writing! :)
Thanks Charity. And you're right. They weren't gonna read them anyway. Better using your energy on those who would.
Yeah. You don't want it to just rewrite you. Better if you go by section and work with it as a collaborative editor. I'd be careful with research due to hallucinations if it's looking at online databases. Working with customs can also help depending on how you build them. I use one to fact check me to make sure I'm accurate in what I write about. A different custom helps me edit stylistically where nuance and narrative style matters more. But yeah. Can be very useful. All depends on the user.