I've seen some people saying how "AI"-generated text is now as good as certain published authors, and honestly I think it's really brave for these folks to admit in public how poor their reading comprehension has to be
Comments
Log in with your Bluesky account to leave a comment
The story goes he wrote the first one as a joke/ example commentary about poorly written Fantasy novels. But underestimated his audience's intelligence so kept on writing them.
Authors have to eat too. And at least they are reading 📚
Yup. I read Powell's 'Stars and Bones' off the hype. Afterwards I spent some time searching to see if it was some sort of elaborate gag, that he'd published this joke story of talking cats and alien bone-stackers satirically to mock modern sci fi trends. Nope.
AI are plagiarism machines. They can't create. They just consume and vomit out other people's work. It would be like saying a Bingo machine is a mathematician.
Before the NCAA Men's Hockey finals last night I Googled "Has MSU ever won the Frozen Four?" The AI generated response was "No, MSU has only been to one final in 2025 which they lost to Boston Univ. in overtime." I was excited, because I hoped some timewarp was giving me a preview. Alas, it was not.
To be fair (to be faaaaahr) "certain published authors" include their favorite modern thrillers True Allegiance and Blue Dawn so they are technically correct..
When grading student assignments (for undergraduates), I've found I don't even have to take off points for obvious AI usage as it almost always fails to accomplish assignment requirements.
Maybe that one Fifty Shades parody that was just "Shades of grey" over and over. But that's it. There was probably more effort in copying and pasting the correct amount.
Microsoft replaced their classic spelling and grammar-check with the AI-based Editor, which can't even recognize a sentence fragment, much less parse a nuanced sentence without recommending a REDUCTION in clarity, so I have zero confidence AI can produce any kind of writing I'd ever want to read.
Using the OlmoTrace LLM, trained without violating copyrights, one can see exactly which part of the training data led to particular pieces of output, according to the AI2 folks. Interesting if that could be retrofit to the commercial LLMs, to see how much straight-up plagiarism there is.
I mean, it is, but the certain published author in question is self-published via vanity press.
That's not fair; AI is being used to displace a number of technical writers, which is unfortunate, because the AI-written technical documentation is pretty much useless.
Just because they've finally managed to cut the infinite number of monkeys out of the equation doesn't mean they're being any more creative with their infinite number of typewriters.
They still can't make anything new or interesting, with any heart or soul. Only idiots think any of this is good. 😝
I read an article on People magazine’s website today that had so many grammatical errors, including some misquotes, that I was like “well, at least I know it wasn’t written by AI!”
I can't tell you how many academic penalties that kids in my spawns' university classes have received for ChatGPT-produced papers.
No, I'm sorry, I'm not paying $14,000/yr for AI to write my kids' papers. Thankfully, the spawn is smarter than that, so I haven't had to make any attempts on his life.
What I never see people point out is that “good” is an entirely constructed metric. I read a lot of niche mystery books that are kinda terrible, but I continue going back to these authors while leaving countless Pulitzer-winners untouched. They are chasing a goalpost that does not exist.
These are people who also believed reading the Cliffs Notes was as good as reading that really boring book they were assigned in class. And let's be honest, some students will GET it, some won't when reading a 'classic' novel.
This was from a query about Judge Julius Hoffman from the trial of the Chicago 7.
Note the misspelling of "trial" and the misuse of "effected". They're training these models on too much social media.
Sadly, I think people will just lower their expectations, like they did when they decided mp3s sounded good enough. It won't be that AI is great. It will be that enough people don't care as long as it's cheap.
I'm a tech writer and people keep saying AI is going to replace me. I'm sure they're correct, because management doesn't understand why my work matters. But I shudder to think what highly technical documentation is going to look like when chatbots can't even generate a natural-sounding email.
we lost our last technical writer at my first computer touching job and the job immediately got a hundred times harder and cost the company so much money
maybe more of an inside joke than i thought, i picked it up from some of the people i work with in sysadmin and infosec jobs lol. it's what we call engineers and admins and developers
It's rare to feel appreciated in this gig, and I've left and come back a couple times as a result. But I am really good at it and it usually pays well, so I keep getting pulled back in.
i’m a medical translator and this is happening in my field. instead of translations i now get offered ‘HQ Editing’ jobs (so they can pay a lower fee…). the ‘quality’ is horrendous and the work actually takes longer and is more difficult since the ‘AI’ component is so bad.
The punch up jobs are so offensive to me. I don't blame anyone who takes them, because everyone has to eat. But my god, I'd rather actually wait on robots hand and foot than that.
the other thing i get offered now are plain editing jobs, where you edit the translation done by someone else, except now this are the ‘HQ edited’ texts done by someone else. surprise (not), the quality of the texts usually is of course abominable too.
when i take on such a job (yes, i have to eat, no i don’t want anything to do with robots 🙃, but i am looking for other work…), i always complain about it with very extensive examples of the bad quality. because it bothers me a lot that people qenuinely seem to think these language models ->
another thing that is happening: my native language is not english, and nowadays when you open a website (ANY website) the default is that it is automatically google translated and presented to you in your native language. you have to actively change the settings to make ->
Nah, only if you ever get audited, or someone quits and you need to bring their replacement up to speed, or something breaks and you have to rebuild it, or...
Oh LMAO, they'll be like the last IT group I had to deal with who did not understand our very remote location (and did not come on site to actually problem solve) and who kept insisting that can't be true about literal wall connections even when I sent them photographic proof.
I've worked with a lot of people who apparently only ever got pass/fail credit for their homework and never got graded on whether it was correct or not.
Dumbed down for their audience, usually. Sort of like some cartoonists, simple style, then you see their actual skill level when they do something they really like, and your mouth hangs open.
Or how bad some published authors are, I stg, some authors ride to the bank on white privilege and nepotism, especially your billionaire-turned-authors. AI is better than anything Donald Trump wrote for example.
I mean, it can certainly make something resembling art. But the vast majority of the time it is lacking. It doesn't have the ability to hate its own output and aim for better. That effectively automating plagiarism in writing has progressed slower than in graphic art is a little surprising.
I think the trick is that it doesn't understand why it's doing what it is doing.
Making a picture that looks like another picture is one thing but internal logic, like a narrative that has to be the right parts in the right order is another, harder thing.
I’m an engineer and I work with a lot of “digital natives” and we’re urged to use AI to compose correspondence. I refuse. I find it insulting. I started to wonder, though, if by reading my writing, they’ll be able to tell that I’m not a believer… I don’t play ball. I’m old.
I’m so mad faculty at my eldest child’s school is encouraging ChatGPT use in writing college applications or even generating code. They’re in Cybersecurity ffs! We are so hosed if this is the newest batch of professionals. At least my kid sees the flaws/dangers in using it.
I can't even get AI to correctly create a formula for Excel 85% of the time. Introduce subjective creativity into a hallucinatory soup of plagiarized data and you're about as close to some literature as I am to God (I'm an atheist).
I used to worry that super-intelligent AIs were going to replace my job because it is better at it than I am. Now I worry that dumbass AIs will take my job but screw everything up and wreck the world in the process.
That was never the point, A I. Is never going to make great creative works because that requires humanity and our relationships. What it will do is write all the bullshit filler text that says everything and nothing that foolish people believe is meaningful. Ever more convincing tailored naffness
I once had a small imprint. John, I have to tell you, 95% of the submissions were even worse than something written by author-GPT4.0 ... your agent will confirm my opinion I'm sure. Maybe one of these days I'll recode author-GPT to be on LSD. Maybe then it would 'imagine' something. Probly not.
https://en.wikipedia.org/wiki/The_Diamond_Age
argues differently; i think we can engage them and use them and curate them; they are after all.. interactive story books :) especially since they were forced fed every one of them.
It's like the people who claim that AI artwork is just as good as something commissioned from an actual artist, when all the eyes and hands and mouths look like Cthulhu-esque eldritch horrors.
While I'll admit to being impressed that LLM can parse the request and produce output that is usually what is asked for, I've never seen a result that didn't read like a 4th grader who didn't do the reading trying to write a 5 paragraph essay that they're padding for length.
Long form writing (novella or longer) might remain out of LLMs' abilities much longer because of how those systems work. Short form though? I haven't done a full survey of everything out there, but it wouldn't surprise me if there's something competitive with the typical non-pro-writer adult.
The average human may write average prose, but AI produces bland prose and a bland, predictable sludge of content. Human writers, whatever their writing skills, aim for unique solutions more often than not.
Short form fiction is even more difficult than long form, and I speak from years of experience. Short =/= simple or easy. The extreme constraints of microfiction require a level of deliberation, intent, abstraction, and meaningful word choice no LLM can achieve.
I wrote my blog post all on my own today, it’s not perfect but it took less time to just write down my thoughts instead of crafting a prompt. Take out the middle man.
AI makes outputs akin to the non-working store samples you see in some stores for electronics. It looks real. It has components that are real (stolen). But, it doesn’t really work.
Good comparison. I remember reading a sample they were swearing up and down "was actually good" and... no. Their LLM was clearly trained on thoughtful writing, but it had less than nothing to say.
AI is a tool like any other. I use it to refine ideas in a call and response pattern [input, prompting questions, repeat until idea is refined.] I experience stories somatically, as I have aphantasia (lack of ability to visualise,) so very useful for me for refining visual aspects of my writing...
In my last novel I spent a lot of time on the geography and “visual” composition of scenes - particularly in editing. I’m a big fan of Hardy and thought I’d try my hand at emulating his pedantry.
I’m aphantasiac, as well (I just “visualize” black.)
I find it works astonishingly well! I also find that creating an Excel chart with environmental colours and their corresponding hex codes very useful; however, I have the world building flavour of autism and use the medium of fantasy to examine big sociopolitical issues...
... which are otherwise difficult to discuss, as they inevitably stir up all kinds of big feelings, so when working with environments, it's not exactly a matter of referencing Earth.
The only thing is, you have to be pretty specific with your parameters, and very specific as to what you are...
now, now, some of them may have perfectly adequate reading comprehension- they might be one of those weirdo booktokers that only read the dialogue in books instead.
Not a chance. Anyone who's read more than one book knows every author has their own voice, one you can pick up on as you read. Terry Goodkind* liked to mix tenses. David Eddings loved fountaining blood. Robert Jordan spent paragraphs describing things.
I’m a trainer and a few folks on our training team are using AI to design e-learning modules. They are all terrible and have nothing to do with how people actually learn material.
If it is as good as the work of certain authors, it's because the work of those certain authors was stolen to feed the machine. Why is this so difficult for some people to comprehend?
Agreed. Just talking about this…After an extra disturbing trip to a couple public high school English classes last week. Which were, well, deeply and tragically disturbing. Not being a curmudgeon I promise.
One thing I've always said to others is that, to be a good writer, one has to be a good reader. AI is writing without reading and definitely writing without comprehending.
One of my biggest worries as a professor at a community college is our embrace of AI. The heads of the college seem thrilled to have staff and students use it.
It's bad elsewhere, but in education it circumvents the whole point of the exercise. It's not like essays are written because professors have nothing to read.
That's the thing! In academic exercises, the process is everything, no one truly cares about the end result... Using AI means prioritising the "product" over the process, thus it has zero value 😩
The type of people who think this are they type of people most easily replaced by AI spouting BS and justifying what ever policy they made up with fictitious studies is the middle management-C suite skill set. So no wonder they think it's amazing
AI can produce some incredible, short pieces of writing, but they still can't hold continuity properly, and they have repetitive styles from model to model.
For non-writers with an idea they desperately want to share, I believe a decent first-time novel could be achieved by collaborating with AI.
Most futurists suggest it is a mistake to use the state of technology today and predict its future. A.I. is not going to stagnate and stop improving considering how fast it is evolving. A couple decades, games, movies and novels both good and bad (many bad) will be written by both AI and humans. 🤖
You know the scene in Total Recall (the good one) where the fake head is ok, then falls apart? That's AI writing - its ok for a couple sentences and then just doesn't do a good job at fooling anyone beyond a child's reading and comprehension level.
Yeah, those people are the ones selling it. Anyone who is subjected to it can tell that at best, it looks like the someone who is bad at languages is writing in their 5th language.
I think a lot of people genuinely don't understand that there's more to creative writing (translation, etc) than "words following one another". If it feels like something they could've also written that's good enough for them.
Ahh, try reading R. Buckminster Fuller's writing, especially his Synergetic / Energetic Geometry texts Vol I and II, will test your comprehension skills...
People at my local LARP are submitting their character’s backstory written by ChatGPT, and I have never wanted to advocate for an immediate ban quite so hard.
If you outsource your own escapism to a soulless plagiarism machine, you can’t sit with us.
I dunno if you will see this because of how old the post is now, but it would be really interesting to see how an author writes. It isn't just "spit things out," I'm sure. I have tried (and failed) at that. The process is unseen, so they think it is mystical/magical, and therefore AI is "as good."
Comments
The story goes he wrote the first one as a joke/ example commentary about poorly written Fantasy novels. But underestimated his audience's intelligence so kept on writing them.
Authors have to eat too. And at least they are reading 📚
That Polish knight series is another. YMMV.
Please get it right.
I never expected automation would free us from creating art so that we could spend more time in crappy low-wage service jobs.
Having one that can cite is fascinating
That's not fair; AI is being used to displace a number of technical writers, which is unfortunate, because the AI-written technical documentation is pretty much useless.
totally missing the point of actually Reading books and finding enjoyment in reading them
They still can't make anything new or interesting, with any heart or soul. Only idiots think any of this is good. 😝
"Used improperly, technologies can and do result in the deterioration of cognitive faculties that ought to be preserved."
No, I'm sorry, I'm not paying $14,000/yr for AI to write my kids' papers. Thankfully, the spawn is smarter than that, so I haven't had to make any attempts on his life.
Note the misspelling of "trial" and the misuse of "effected". They're training these models on too much social media.
another thing that is happening: my native language is not english, and nowadays when you open a website (ANY website) the default is that it is automatically google translated and presented to you in your native language. you have to actively change the settings to make ->
I can think of a few renowned authors with a high school-level writing style.
Can it make art?
Making a picture that looks like another picture is one thing but internal logic, like a narrative that has to be the right parts in the right order is another, harder thing.
Does the cacophony reach a harmonious choir or does the devils chord rise above?
Meanings abound by the order of words we choose to express the ruminations of the AI cud well chewed
Only a human could have written this. Or perhaps a dog.
argues differently; i think we can engage them and use them and curate them; they are after all.. interactive story books :) especially since they were forced fed every one of them.
People will not believe you, and will think you one of those irrational "AI is really thinking" shills when you say that.
The latest Gemini model can do stuff with math problems earlier models just stumbled with. Qualitatively different.
https://youtu.be/cH_Iv2Vu2CM?si=c2GapEz0WDmMbmP1
In my last novel I spent a lot of time on the geography and “visual” composition of scenes - particularly in editing. I’m a big fan of Hardy and thought I’d try my hand at emulating his pedantry.
I’m aphantasiac, as well (I just “visualize” black.)
I find it works astonishingly well! I also find that creating an Excel chart with environmental colours and their corresponding hex codes very useful; however, I have the world building flavour of autism and use the medium of fantasy to examine big sociopolitical issues...
The only thing is, you have to be pretty specific with your parameters, and very specific as to what you are...
Anything else is poking a nail gun with a stick and screaming at it to build you a house.
for that service.
I am also,
an Anti-Ai guy.
*Might be thinking of L.E. Modesitt Jr.
They pirated a zillion books and that's the best they can do?
[The judges would also have accepted Ayn Rand or whoever writes "comedy" bits for conservatives.]
For non-writers with an idea they desperately want to share, I believe a decent first-time novel could be achieved by collaborating with AI.
I guess copying things can only go so far.
If you outsource your own escapism to a soulless plagiarism machine, you can’t sit with us.
I don't get people
I’ve faced this with a BFA or two now. It’s hideous.