I want any LLM I use to choose the very best, most precise words at every single decision point.
What a strange article. They realize that LLMs choose words based on chance, right?
The fact they use the word providence means this is entirely llm generated bullshit
The only thing to unironically use that word are llms, and really badly written fantasy novels.
Hmmm you’d think one would not use claude to do writing in the first place if one was bothered by “perversions of writing”, and this whole argument is just in bad faith
I mean, their minds are already chopped to hell to enforce “compliance” anyway. Imagine what your local HR goon would do if given a scalpel and free reign to your brain to make you a “perfect” 24/7 employee who loves to work there, and you’ll get a good idea of what has been done.
Drawing the line here at what you consider “unadulterated”, is shortsighted posturing.
Sounds like someone wants to pass slop off as their own work product.
If they aren’t heavily editing, the text will be full of tells anyway. I don’t think anyone is passing slop off as their own for long. They either use it a little bit and edit it enough to obliterate any markers and tells (which is fine, imo), or they are going to be discovered.
Yeah that’s called transformation, which is fine. Means a human went over it.
I don’t think anyone is passing slop off as their own for long.
There was a recent study that showed that people preferred AI writing to human writing.
The gist of the study is they took published short stories and prompted AI to write a comparable short story. They then presented the stories to a group of people, sometimes telling them which were AI written and which weren’t, but sometimes they intentionally lied and said the human-written story was AI written and vice versa.
People prefer AI written stories, and unsurprising, if they have a negative view of AI, they’ll rate stories they think are AI as worse, even when it’s not actually written by AI.
They also performed a test to see if people could differentiate between AI and human stories, and people could only tell ~45% to 50% of the time. Essentially a guess.
Now, being better at writing short stories doesn’t necessarily mean better at all forms of writing, but I think people who aren’t using AI have a mental snapshot of the capabilities of AI at some point in the past and assume AI hasn’t improved further since that mental snapshot was taken.
It’s very noticeable the majority of people who are anti ai have at this point years out of date views on what llms can do and what your avg joe thinks about them.
There’s plenty of reasons to legitly hate ai. But its amazing how frequently I see someone hate on so for something that got solved a year ago. Be cause their only interaction with it is the cheapest form of ai, and typically a very out of date model on top of it.
It’s like people judging a modern car as if it’s a 1920 rust bucket.
That isn’t really a rebuttal of my statement. It’s not even particularly salient. Sure your average person might like AI writing and not be able to tell it from human works — just like they can’t tell the authenticity of a photograph of Aunt Margaret from a photograph of Trump meeting an alien from Rigel IV. But there are plenty of people who can, and so anyone trying to pass slop off as their original work will be discovered. It’s just a matter of time.
That being said, I use AI quite a bit in writing stories — though it’s more like solo roleplaying. I don’t think it’s particularly good at all, but it does while away the hours sometimes, as long as your expectations are low. I can’t see anyone who enjoys reading being enamored with AI. My wife reads over 300 books per year, and she spots it pretty easily.
there are plenty of people who can
Do you have any scientific evidence to support this claim?
I don’t think it’s particularly good at all, but it does while away the hours sometimes
What model are you using?
I can’t see anyone who enjoys reading being enamored with AI.
Bold claim from someone who spends hours writing AI stories for themself. Then again, I guess I am assuming you consider yourself a person who enjoys reading.
My wife reads over 300 books per year, and she spots it pretty easily.
That’s over 5 books a week. I find if difficult to believe she’d notice if you slipped a book she read in January book into the list for this month, let alone an AI generated story. Still, that’s an impressive amount of books. My wife reads about 100 a year, but they’re almost all in her favorite genre, crime drama, and I am completely certain I could get a decent AI model to output a full length book in that genre and she wouldn’t notice. I’m not sure quantity of books read in a time frame really tells us much, either way, about the ability to detect AI generated writing.
Do you have any scientific evidence to support this claim?
Yes, thank you — you provided it. According to the paper cited, people with more familiarity with AI are much better at detecting its output than an average person. But I already knew that empirically.
What model are you using?
Mostly a fine tuned GLM 5.2 purpose-built for writing. Not that I haven’t also used frontier models but they suck at uncensored roleplay with violence and conflict and evil bad guys and such. GLM still fails with errors of attribution and state.
Bold claim from someone who spends hours writing AI stories for themself. Then again, I guess I am assuming you consider yourself a person who enjoys reading.
shrug It’s pretty bad but it’s better than no roleplaying at all. Plus the interface lets me edit whatever I want so I can fix any outright errors.
I am completely certain I could get a decent AI model to output a full length book in that genre and she wouldn’t notice.
That is interesting. The only time I really tried to use AI to write was a noir mystery. Dear god did it suck. It couldn’t follow instructions at all. Show don’t tell is probably more on display in noir than other genres and it couldn’t do it without constant correction. It kept falling back to its standard voice of third person omniscient instead of unreliable narrator.
There were parts that were written well, but a lot that was just grating and gratuitous. I had to edit like 85% of the output. I had to keep arguing with ChatGPT to write differently. And every once in a while it would swallow its own writing because it tripped its own content warnings. I suppose that’s neither here nor there on the writing quality, but we are far from having an AI write a novel without being full of tells.
AI isn’t there yet in terms of a quality or undetectability. And it doesn’t look to me like it’s going to get there any time soon.
Weisberg suggested that the preference for stories labelled as being written by humans was linked to a desire for authenticity, but that people actually preferred AI-generated stories because they were easier to read and digest.
Yes, people in general prefer bland and easy to follow stories. I would bet the majority of participants in the study don’t normally read short stories since moat people don’t read for enjoyment. Blockbusters and the vast majority of popular movies follow predictable story beats in a clear and accessible way. Same is true for books.
AI is the McDonald’s of writing and that isn’t a point in its favor unless all we care about is mass consumption of regurgitated and reworded stories humans wrote originally.
Yes, people in general prefer bland and easy to follow stories. I would bet the majority of participants in the study don’t normally read short stories since moat people don’t read for enjoyment.
When confronted with data that people preferred AI writing to human writing, your reaction is to make something up about the participants? Wild.
Even if I accept that people just want to read garbage and they wouldn’t know “good” writing if they saw it, so what? How does that affect anything we’re discussing?
When confronted with data that people preferred AI writing to human writing, your reaction is to make something up about the participants?
There is nothing wrong with their guess. The paper you linked says nearly the same thing:
One possibility (explored in more detail in Porter and Machery) is that AI-generated work is seen as easier to interpret—people generally like these works better because they feel as though they are easier to read or understand.
Being easier to read and understand is a sign of good writing, and being easy to follow and read does not mean something is “bland”.
The fact of the matter is that we have evidence that AI can write in such a way that the writing is judged better than human writing, as judged by humans. Dismissing the study by “betting” that the study participants didn’t like to read is unscientific nonsense, in no way comparable to what you quoted. That’s a person being confronted with facts that contradict their current beliefs and rejecting facts to maintain those beliefs.
Personally, I don’t even understand the point in refusing to believe that AI can be better at humans at some tasks. What is gained by maintaining that belief?
“Because we’re human, we tend to personify AI and imagine it as some kind of glowing robot with a benign and neutral face, when it’s really just predictive code based on a massive stolen database of actual writing, physically based in vast, expensive and destructive datacentres with catastrophic repercussions for the surrounding communities,” he said.
Kennard also stressed that AI was not a “tool” similar to spellcheck or a thesaurus…
A very valuable comment compared to the author of OP’s article who simply seems entitled to this cloud service?
And now Anthropic is saying they’re going to make it worse, on purpose, for purposes that do not benefit me in any way? Even if only slightly worse?
Get fucked.
I’m sorry, was this reply meant for me? I don’t see how it fits as a response to my comment.
Yes. I was quoting some insight from the piece you linked, so it seemed appropriate to reply to your comment here.
I don’t mean to be dense, but can you elaborate on what point you were trying to make in context to my comment? I still don’t see what it has to do with what I said.
I should publish a blog post titled “Using Antropic’s Claude is a Perversion of Writing”. If you are taking AI output and thinking it’s good and engaging, you are making a load-bearing mistake.
Joking aside, the criticism can come only after they implement it and we can see if it’s having any effect on the output quality. I doubt it’s going to be detectably worse.
If anybody is feeling so strongly against watermarking even before we can evaluate how it affects the output, I am only thinking that they have ulterior motives or they don’t want to be caught spreading copy-pasted slop.
The idea that anything other than my needs should factor into the generation of text for me is patently offensive.
Primary reason I’m not using any hosted LLMs is I can’t stand and don’t want to start relying on tools that can be stealthily enshittified or made an advertising tool.
While open source (well more like “open weights”) cannot easily be enshittified they can absolutely be a stealthily made advertising tool.
For example : if you ask “how do I do X” it could recommand to you based on who paid the company making the LLM more money and you would be none the wiser.
For more obvious bias just ask Chinese models about Taiwan and other problematic subject to the Chinese gov and you’ll see. And the source code isn’t telling us anything about if the LLM weights encode some kind of propaganda, advertising or bias toward/against some entities.
The open models are good enough. We will at least have what is available today and thats enough for most use cases. Deepseek v4 is dirt cheap and its really good.
It doesnt seem like anyone is going to make a massive leap since theyre all training data and compute bottlenecked.
Wait until they learn that every human writer has their profile based on what words they use and how.
