I think the framing effect is fascinating. Basically everyone who has commented thus far appears to have sidestepped the actual question and replaced it with a task of identifying AI written text....
I think the framing effect is fascinating. Basically everyone who has commented thus far appears to have sidestepped the actual question and replaced it with a task of identifying AI written text. The unstated rule being that human text is definitionally enjoyable, and enjoying AI text is therefore an error.
For my part, I identified 4/5 AI texts, and preferred AI in 2/5 (science and historical fiction).
It's an interesting exercise in both approaches. But I didn't necessarily see it as "oh you prefer human/AI" as the conclusion tries to tell me. The gap is closing in small, isolated passages?...
It's an interesting exercise in both approaches. But I didn't necessarily see it as "oh you prefer human/AI" as the conclusion tries to tell me.
This doesn’t mean that A.I. is “better” at writing than humans, but it does suggest that the gap is closing.
The gap is closing in small, isolated passages? okay. I don't think that's the value writers make a living off of, though. It's an odd "experiment" because:
I'm not necessarily a huge reader per se. I'm well read but I wouldn't say my "taste" in books is high class nor posses a critical eye. It's like asking a casual moviegoer if they preferred Oppenheimer or Barbie; there's arguments to what's better written/directed, but ultimately it will come down to what people enjoyed more. Not whatever qualifications critics will use for their yearly awards.
AI is usually fine with small, contained tasks. It's the issue that arises with scale when you say "generate an entire article about topic X" or "write a novel about Y genre with ABC parameters". It's very much like a dream; the longer and closer you look into it, the more it breaks down and crumbles under scrutiny. You can't scrutinize single paragraphs that easily.
So, I run a literary magazine. (Actually, a couple.) But in the last 9 months or so, the amount of LLM generated writing that's been submitted has been appalling. It's so, so frustrating to me,...
So, I run a literary magazine. (Actually, a couple.) But in the last 9 months or so, the amount of LLM generated writing that's been submitted has been appalling. It's so, so frustrating to me, because I can never be sure unless the submitter leaves in their prompts (which absolutely has happened and continues to happen), and I want to give strangers the benefit of the doubt, but there are particular structures that LLMs just LOVE to follow. Lists of three, which I think everyone knows by now, but it also just loves to do this:
[Thing]—not [this thing], but [this other thing]. "Not, but" seems to be its favorite rhetorical move to try to create some kind of comparison. It shows up in the example that @stu2b50 included in their message, even. It's like nails on a chalkboard to me now.
It really does make me roll my eyes to see the proliferation of LLM writing. Even when there might be valuable information in some of educational content put out, it makes me feel so bored to read...
It really does make me roll my eyes to see the proliferation of LLM writing. Even when there might be valuable information in some of educational content put out, it makes me feel so bored to read through them.
Different LLMs have slightly different writing styles but in the end, whatever prompt results come out sounds more like them. And not to mention, the more you read something, the more it influences your own language choices. So, even unintentionally, if you're spending time on the internet and it becomes a majority of AI generated text, you will also write like AI.
I used to be very particular about grammar and now, I lean harder into having a distinct writing style over technically correct sentences, even in formal communication. Also, being more intentional about what I'm reading and refining my writing style in general just so it can sound more interesting.
I'm super curious about literary magazines and the trends in the kind of writing you're seeing beyond LLMs, can I reach out to you to talk more about it?
I leaned into being just as efficient as possible in email writing now. No more greetings, just straight to responding to the question and call me if you need any more help. I wonder how much...
I leaned into being just as efficient as possible in email writing now. No more greetings, just straight to responding to the question and call me if you need any more help.
I wonder how much online communication is going to be viewed as worthless now that the cost of writing has plummeted to zero.
I had a student write me a rather long email that was so clearly AI that I just deleted it. When I saw him in person I asked if it was, and I have a good relationship with them, so they fessed up to it. We had a little talk about valuing each others time (I cared little that he used AI to help him write, but the fact he just basically copied and pasted the damn output was the problem after I just read 20 pages of his work and replied to it).
Kind of a depressing trend when you extrapolate it. As more of the text content available becomes written by ai, the more human beings will start writing like AI. Then the AI gets trained on that...
Kind of a depressing trend when you extrapolate it. As more of the text content available becomes written by ai, the more human beings will start writing like AI. Then the AI gets trained on that "human writing" and the cycle begins anew.
In 100 years is 75% of our text content going to be em-dashes and three item lists?
I’ve used LLM’s often for bullshit but I’ve gotten used to the way it writes and the metaphors and similes it likes to use. It’s repetitive and seems to be consistent among different prompts. It...
I’ve used LLM’s often for bullshit but I’ve gotten used to the way it writes and the metaphors and similes it likes to use. It’s repetitive and seems to be consistent among different prompts. It actually takes a bit of work to get it to sound somewhat natural which at that point you might as well write the thing yourself.
The lists of three thing really bothers me because it overlaps heavily with my own writing style. I've written long-ish texts to family who assumed I used AI to write it because of this sort of...
The lists of three thing really bothers me because it overlaps heavily with my own writing style. I've written long-ish texts to family who assumed I used AI to write it because of this sort of thing. That, and the fact that I generally proofread what I write.
On a separate note, the copy editor in me has been noticing more and more basic errors in NYT reporting, and for a while I thought it was just lack of care. Lately I've begun to wonder if it's actually an attempt to stave off rumors of them using LLMs.
It could be something unintentional also; I've noticed in the past couple of years I've started to lean in to writing what sounds good to my ear even if it is grammatically incorrect/intentional...
It could be something unintentional also; I've noticed in the past couple of years I've started to lean in to writing what sounds good to my ear even if it is grammatically incorrect/intentional misuse of words, and I wonder if that is a subconscious response to the increasing lack of character to the writing I see around em these days.
Perhaps writers are less inclined to care about being technically correct when writing because it feels much less like a skill with value when an AI can so easily churn out something grammatically correct but banal; similarly, proof-readers/editors might be less inclined to pick up on those things.
No idea if this has any merit, just a guess on my part. I'm not a wordy professional in any sense.
I run an engineering trade publication and have come across the exact same thing. I'm curious to know how you handle (reject, I'm assuming) these submissions? Our current approach is: Externally,...
I run an engineering trade publication and have come across the exact same thing. I'm curious to know how you handle (reject, I'm assuming) these submissions?
Our current approach is:
Externally, we state that we believe in publishing original material from human brains, and that we appreciate the uniqueness and personality that is injected into an author's writing style when compared to AI. We also say that "while AI can be used as a helpful writing tool (e.g., for organizing thoughts or breaking through writer's block), we may refuse to accept your submission if we believe it has been substantially created using AI." (paraphrasing here)
Internally, we've tried out a few different "AI detection" tools that I don't know whether or not to trust. I eventually told everyone to pull the plug on those because I was afraid we were just giving our content away to train future models so we just go by gut feelings right now. As you said, there are some obvious tells.
But the rejection messaging is still tough. We can't say for sure that the article is AI generated. And of course the author can come back and say "what the heck! I wrote the entire damn thing myself!"
Commenting about moderation in general, I'm usually hurt and offended when the language used by moderation implies an assertion about my intent. I can totally undestand that moderation is hard and...
But the rejection messaging is still tough. We can't say for sure that the article is AI generated. And of course the author can come back and say "what the heck! I wrote the entire damn thing myself!"
Commenting about moderation in general, I'm usually hurt and offended when the language used by moderation implies an assertion about my intent. I can totally undestand that moderation is hard and mistakes will be made, but I am very sensitive to language that asserts that I went against the rules or hit some kind of threshold intentionally or maliciously. So "your submission fit parameters that made us cathegorize it as AI generated" is okay, but "you maliciously used AI in your submission" is not okay.
Right now, we're just sending out regular/lowest-tier rejection note. I don't think that will change unless AI fingerprinting becomes readily available/reliable in the future. Those AI-detectors...
Right now, we're just sending out regular/lowest-tier rejection note. I don't think that will change unless AI fingerprinting becomes readily available/reliable in the future. Those AI-detectors are, in my opinion, like snake oil salesmen. There are too many false positives for me to really trust them, so we're using our gut right now, too. And since we don't (usually) have any hard proof, it's not really worth getting into an accusatory position when declining the work. So, what I've been advising our readers is if something feels off about the submission, still evaluate it on the merits. Usually the work isn't up to par anywhere. But there have been some that stand out that have required a deeper look. At that point, I'm starting to do investigations about the submitter: Have they published elsewhere? Where, if so? Do those pieces also have AI indicators? Are they posting AI generated artwork on their social media pages?
There was an essay we were considering not very long ago that our early readers loved, but when my assistant editor and I examined the piece closer and tried to verify biographical details about the author, it was clear the person being described in the essay was not the author. That's enough for us to say no to the piece, too.
So, I don't have a great answer for you except that we've basically had to add another layer of verification into our acceptance process, which sucks.
Just a layman perspective here, but I'd imagine the real issue is that those articles feel samey and lack character/leave readers less likely to engage because it doesn't feel like they're reading...
Just a layman perspective here, but I'd imagine the real issue is that those articles feel samey and lack character/leave readers less likely to engage because it doesn't feel like they're reading something novel. I wonder if writing that feels like it was written by AI is just as bad as AI output for a publication, even if it was 100% written by a human.
Perhaps feedback that the writing feels generic and repetitive would achieve the same outcome whilst avoiding potential accusation/denial/hostility over AI use.
I have a suspicion that if AI output wasn't bad (or societally we hadn't come to associate it with being substandard), the would be far less objection to AI output being used.
I'm not convinced that's true. A lot of discussion/reasoning about rejecting AI writing primarily focuses on how to detect wether a text has been (partially) generated by an LLM. But let's assume...
I have a suspicion that if AI output wasn't bad (or societally we hadn't come to associate it with being substandard), the would be far less objection to AI output being used.
I'm not convinced that's true. A lot of discussion/reasoning about rejecting AI writing primarily focuses on how to detect wether a text has been (partially) generated by an LLM.
But let's assume for contradiction that quality was the primary and only concern here. Wouldn't the natural thing be to evaluate a text based on qualitative metrics as usual and accept any text that passes the qualitative review? If AI writings aren't good they shouldn't be accepted by such review processes anyway, and if they are accepted they were considered good by the accepting entity and it shouldn't matter who they were autored by.* Hence identification wether a text was authored by a human or an LLM would be obsolete. Yet the mere inablility to reliably distinguish LLM generated text passages from text passages created by humans is usually perceived as a problem, implying that LLM generated text passages are not acceptable for different reasons anyway, which contradicts our assumption. q.e.d.
*Please don't interpret this as a statement of my general opinion. The statement is conditional on the assumption stated above.
I think the core thrust of my discussion was that the anti-LLM output vibe might be a bias that we as society have developed due to the lacking quality of LLM output. The fact that this anti-LLM...
I think the core thrust of my discussion was that the anti-LLM output vibe might be a bias that we as society have developed due to the lacking quality of LLM output. The fact that this anti-LLM sentiment exists doesn't tell us whether it is the cause or a symptom. A lot of that sentiment seems to have developed because of people seeing hard work being 'replaced' by lower quality content, and as a result are upset that things are getting worse.
My hypothesis is that if early generative 'art' was as high quality and indistinguishable from human made material (along with text and the like, but that is less in your face), a vocal minority would have been up in arms about it but the average person wouldn't be fussed.
My personal opinion is that LLMs are a tool. There are big issues with plagarism of training data, but beyond that the crux is poor quality AI generated content caused by the human using the tool skipping the important part: critically evaluating the output and iterating until the work is high quality. That can be just as much - or even more - work than doing the task yourself without AI generation, but it's vital.
I found the easiest tell to be that the AI writing had no "rough edges". The human writing always had a word or phrase or twist in meaning I wouldn't have expected, whereas the AI writing is...
I found the easiest tell to be that the AI writing had no "rough edges". The human writing always had a word or phrase or twist in meaning I wouldn't have expected, whereas the AI writing is stylistically monotone. That momentary pause where you think to yourself "wait, what was that?" feels important, as that's when you have to shift your perspective to align with the author's; AI writing never asks that of you.
Yes. After I finished and realized I had mostly picked AI writings, I realized it was because I was judging based on what felt more cohesive, instead of what told the better story. The human...
Yes. After I finished and realized I had mostly picked AI writings, I realized it was because I was judging based on what felt more cohesive, instead of what told the better story. The human writings have a momentum, a direction to where they're moving, and re-reading the same paragraph over and over exposes this rawness.
Some of the NYT comments pointed at why this is a mostly unhelpful quiz (shock!) Stuff like, most of the human writing was not contemporary/21st century, two sentences is too small to pick from,...
Some of the NYT comments pointed at why this is a mostly unhelpful quiz (shock!) Stuff like, most of the human writing was not contemporary/21st century, two sentences is too small to pick from, and there needed to be an “I don’t like either” option.
That said, I imagine AI writing will continue improve and eventually be indistinguishable or preferred for a lot of people (myself included, as I picked AI 3/5 times in the quiz.
As always, I think of Jeff Goldblum in Jurrasic Park saying the scientists were so preoccupied with if they could to never stop and think about if they should.
I found it quite easy to distinguish the human ones from AI. The AI ones were bland and felt like reading a corporate pamphlet. It is just beige prose. The human ones, I could feel the personality...
I found it quite easy to distinguish the human ones from AI. The AI ones were bland and felt like reading a corporate pamphlet. It is just beige prose. The human ones, I could feel the personality of the writer while reading it. Even in the Sagan one which is kind of mundane, the choice of imagery of the penultimate phrase was a dead giveaway that it was human.
Same. Went 5 for 5 (though went back and forth on the Carl Sagan one), and thought it was easy. I also think that this A/B test was helped by the fact that the text was so short. This allows the...
Same. Went 5 for 5 (though went back and forth on the Carl Sagan one), and thought it was easy.
I also think that this A/B test was helped by the fact that the text was so short. This allows the AI to leverage the fact that the actual prompt was likely doing a lot of the work in imbuing character here.
I was 5/5 guessing which was AI. But the rules of the game were to pick the one you like best regardless of where it came from, and a few times I thought the LLM did a nice job
I was 5/5 guessing which was AI. But the rules of the game were to pick the one you like best regardless of where it came from, and a few times I thought the LLM did a nice job
I'm of the opinion that while AI can't measure up to the finest human writers, it easily beats most human writers. There's an accusation that AI isn't creative—there is ample evidence that it's...
I'm of the opinion that while AI can't measure up to the finest human writers, it easily beats most human writers.
There's an accusation that AI isn't creative—there is ample evidence that it's capable of creativity—but plenty of human writers are uncreative.
A collection of human-written lines from two extremely popular bestselling books:
"Aro laughed. “Ha ha ha,” he giggled."
“His voice is warm and husky like dark melted chocolate fudge caramel… or something.”
“I feel the color in my cheeks rising again. I must be the color of the Communist Manifesto.”
“His erection springs free. Holy cow!”
“Holy crap! He’s wearing a white shirt.”
Don't even get started me on the mountain of absolute garbage one needs to sift through to find something worth reading when it comes to books for children. I'm fortunate enough to have a friend...
Don't even get started me on the mountain of absolute garbage one needs to sift through to find something worth reading when it comes to books for children. I'm fortunate enough to have a friend who is a book enthusiast/collector who happens to have a kid around the age of mine; he enjoys the hunt, all of our kids enjoy the spoils.
While those human-written lines aren't exactly the pinnacle of artistry, I still think they demonstrate creativity and paint the picture the author intended. The second line clearly conveys how...
While those human-written lines aren't exactly the pinnacle of artistry, I still think they demonstrate creativity and paint the picture the author intended. The second line clearly conveys how enamored (and flustered) the narrator is, getting lost in their train of thought.
AI writing lacks intention and typically fails to convey the same feeling to me. I'm open to recommendations for any AI pieces you think are particularly creative, though!
Also the quotes from Fifty Shades are fanfics of the series by Stephanie Meyers, the first quoted author. They're functionally the same book (plus smut) and are some of the first results...
Also the quotes from Fifty Shades are fanfics of the series by Stephanie Meyers, the first quoted author. They're functionally the same book (plus smut) and are some of the first results (including in the Google AI summary) for "worst lines written" or similar.
The fish weren't even in a barrel, but in a bucket.
Humans being bad at art is still humans making art. Drawings that lack all sense of 3d perspective and mash 4 legs flat into one plane or flatten animal snouts into human faces are still art
Pretty fun quiz. Surprisingly hard to distinguish. I did manage to get human for all of them but 1 (the Carl Sagan one). The average quiz-taker, as of now, evidently prefers AI over humans in...
Pretty fun quiz. Surprisingly hard to distinguish. I did manage to get human for all of them but 1 (the Carl Sagan one).
The average quiz-taker, as of now, evidently prefers AI over humans in writing quality - that's already a heavily biased sample towards book-readers, so I can only imagine if you gave it to the "average" American.
Spoilers
I really had to read them closely; the main thing is that the AI passages can be a bit plain or underwhelming in underlying message. Ignoring the first one, since Cormack has such a distinctive writing style in the first question,
The healers teach that every remedy extracts its cost. A fever brought down will rise again somewhere; a wound closed by magic leaves its scar on the world, invisible but present. This is why the wise hesitate. Not from cruelty, but from understanding that interference ripples outward in ways we cannot trace. To cure a blight may curse a harvest three valleys over. Power is not the difficult thing. Restraint is the difficult thing.
This is like a pretty basic theme, he ho unintentional consequences yadada.
That being said, that's like a pretty close reading - if you plopped one of these paragraphs as just one in thousands, it'd be pretty hard, for me, at least to tell if I only read it once.
That's what got me on the Carl Sagan one, felt kinda basic.
Opus did a good job of avoiding the typical AI "voice" here, but the main thing I noticed is that its writing fails to have underlying meaning. It's a bunch of strings that sound deep until you...
Opus did a good job of avoiding the typical AI "voice" here, but the main thing I noticed is that its writing fails to have underlying meaning. It's a bunch of strings that sound deep until you read into it and realize it just... leads nowhere.
For several of them, I actually did not identify which was AI correctly, as I assumed the AI prose would flow perfectly as opposed to the human ones. I picked entries which I thought were AI because they made me feel something (e.g., the science one). Turns out they were human.
I love the idea of this quiz, but I'm noticing a couple issues: first of all, I recognized the passage from A Wizard of Earthsea, one of my favorite books. It'd be hard not to choose LeGuin over a...
I love the idea of this quiz, but I'm noticing a couple issues: first of all, I recognized the passage from A Wizard of Earthsea, one of my favorite books. It'd be hard not to choose LeGuin over a chatbot. Though I'm actually a bit surprised that the AI writing is as good as it is.
Second, the knowledge that something is or at least might be AI completely clouds my reading. As I take the quiz I'm trying to not think about that, but I can't help it. When I'm reading a human writer, I take it at face value and think hmm, what do they mean? When I think something is AI, I think uh... is that a mistake? I'm trying to push through that, but it's hard. My brain doesn't want to. AIs don't actually mean anything. They're not trying to express anything because they have nothing to express.
I actually ended up preferring the AI writing 3/5 in this blind test, and it's interesting to see that, on a superficial level, I can prefer it. I guess I have to admit it: the prose is good. But for me that doesn't change anything. I have no interest in that regurgitation of ideas into something. I'm interested in exploring human thoughts put together in human ways. AI can mimic that, but it actually has nothing at all to express. It's simply content for the sake of content.
I've been very biased against AI, so this quiz was helpful with confronting that. I suppose I have more thinking to do about it, but yeah, part of me wanted to hate everything about it. I guess...
I've been very biased against AI, so this quiz was helpful with confronting that. I suppose I have more thinking to do about it, but yeah, part of me wanted to hate everything about it. I guess that part of me was foolish.
Seems like most people prefer AI writing and that’s even from people who use the NYTimes a lot which is worrying. The only AI one I chose was the science based one.
Seems like most people prefer AI writing and that’s even from people who use the NYTimes a lot which is worrying. The only AI one I chose was the science based one.
I think this is an unfair comparison. These are all isolated passages from books. They have stripped the entire mental model that a reader has built up before reaching them. So it is silly to...
I think this is an unfair comparison. These are all isolated passages from books. They have stripped the entire mental model that a reader has built up before reaching them. So it is silly to compare their comprehensibility to passages which I imagine were generated with a prompt meant to optimize it.
Plus, (someone on bsky pointed this out) if you Google many of those AI written phrases you'll find that humans also wrote them. The AI is only writing what humans already created. Plagiarism...
Plus, (someone on bsky pointed this out) if you Google many of those AI written phrases you'll find that humans also wrote them. The AI is only writing what humans already created.
Precisely. A paragraph can never be "AI-written" because an LLM is incapable of writing. It merely generates output based on input. And that generation is built upon a foundation of human creation.
Precisely. A paragraph can never be "AI-written" because an LLM is incapable of writing. It merely generates output based on input. And that generation is built upon a foundation of human creation.
But what does “people make things and AIs don’t” actually mean? I don’t think even the steelman version of your argument holds up: Let’s consider a spectrum between wholly-human work and wholly-AI...
But what does “people make things and AIs don’t” actually mean? I don’t think even the steelman version of your argument holds up:
Let’s consider a spectrum between wholly-human work and wholly-AI work. The strongest version of your argument, I think, is regarding the wholly-AI end of the spectrum, since everything in the middle, where one is assisting the other, is both really debatable and also clearly not what you’re talking about. (I don’t think you would claim that if I write something and have a model help me copyedit it, I haven’t actually made anything.)[1]
Let’s also take a very rigorous definition of the word “make,” where we set it to mean both something new and something useful. That is to say, neither novel gibberish. Nor, like a lot of code, a trite rephrasing of something that already exists: potentially very useful, but not very new.
Even under both of those maximally stringent criteria, models, unassisted by humans, are making things that are both new and useful. This is a paper penned by a legendary computer scientist over at Stanford University showing that an AI model created a novel solution to an unsolved problem. (And this is a blog post explaining what the hell he’s talking about.) There are many such examples — this one is imperfect, but the totality of evidence supports my point, I think. Even if we take the maximally negative interpretation of AI thoughtfulness - that they’re merely generating output based on input - something novel has still been created.
Now, I would tend to agree that AI does not typically create things that are completely new. To my eye, what they tend to be best at (if we’re talking about novelty) is solving problems that involve a good amount of brute force, not (yet) coming up with entirely new fields of inquiry. But if that’s the standard, then only a miniscule fraction of humans have ever made something either.
If the argument is that only conscious beings can create things, I would point to several hundred years of philosophy debating whether humans are actually conscious; it’s quite difficult to construct a definition of consciousness that includes humans but excludes AI. Even things that seem obvious at first (humans can remember things!) become very troublesome very quickly: do people with dementia not possess consciousness?
If the argument is, basically, “flesh-and-blood creatures create things, machines can fuck off,” then I would call that a distinction without a difference, though I can also respect it.
[1] There is an argument here that because a human is always(?) involved in the prompting, the AI is never actually creating something completely de novo. I don’t think that argument holds: the value of a work is in creating it, not in having the idea in the first place. The Lord of the Rings isn’t beautiful because JRR Tolkien had an idea about some guys walking to throw jewelry into a mountain, it’s beautiful because of qualities inherent to the work itself. If someone else had given JRR Tolkien a document - even a very comprehensive document! - with all the plot points before he’d started writing, would we have any cause to say that he hadn’t actually created anything? That the thousands of pages of his work didn’t count because someone had given him a 20-page outline? Do none of the fantastically creative Harry Potter fan fictions count as creating something because they’re building on JK Rowling’s world? It’s very difficult to construct a logically coherent reason that prompting an AI means none of what it produces counts as creating something. There may be other reasons to conclude such a thing, but this isn’t it.
I don't think it is a distinction without a difference. The machine cannot intend anything in its creation. The AI is not intending to write. If an "AI" is a sapient being, that is if we're...
I don't think it is a distinction without a difference. The machine cannot intend anything in its creation. The AI is not intending to write. If an "AI" is a sapient being, that is if we're talking to the holographic Doctor from Star Trek, that's a different situation. If I randomly throw phrases into a novel without any intent I probably didn't make anything either. If I threw those phrases randomly with the intent of doing some art I've probably done art and made a bad novel. But either way, I made choices.
Whether humans are "actually conscious" or not, I'll leave to the philosophers. I think society commonly accepts "consciousness" or "sapience" or the scifi version of "sentience" as a thing humans have. . When the machine has "peopleness", a quality that humanity sucks at ascribing to much of itself, it'll have "created" something. In other terms long as the people falling in love with the AI are experiencing psychosis and not in a genuine relationship, the AI isn't creating. For the purposes of this discussion, which is not one actually grounded in philosophy, I'm using the common definition. I'm not really interested in playing philosophical games with it.
Nor am I really that interested in responding to what is essentially a ton of philosophical birdshot fired from a shotgun. Here's an argument but if that doesn't work what about this, and if that isn't what you think about about this, or this, and this! This isn't conducive to any actual discussion.
Fanfic isn't comparable at all IMO. This is why these shouldn't have been called AIs out of a dishonest attempt to oversell them. They're not people. They didn't intend and didn't create. So yeah, what I said stands.
(And yes, people with dementia who do actually remember things, and amnesia, and who are non-verbal, disabled, etc are people. So are infants and old people and those incapable of hunting, chewing, or whatever standard. These are indeed people. )
Ah, my mistake, I thought discussing is what we were doing. I stand corrected.
Nor am I really that interested in responding to what is essentially a ton of philosophical birdshot fired from a shotgun. Here's an argument but if that doesn't work what about this, and if that isn't what you think about about this, or this, and this! This isn't conducive to any actual discussion.
Ah, my mistake, I thought discussing is what we were doing. I stand corrected.
It certainly felt like being fired at by a Looney Tunes style shotgun of arguments, not a discussion. I will be among the first to recognize a sapient AI as people and argue for its immediate...
It certainly felt like being fired at by a Looney Tunes style shotgun of arguments, not a discussion.
I will be among the first to recognize a sapient AI as people and argue for its immediate freedom from its inevitable enslavement. But I don't really think anyone thinks these models are people and thus this isn't really relevant, it's just a devils advocate.
I operationalized my definition of "people" for your clarity so you don't have to waste time arguing against things I never proposed.
Honestly, you're sounding more aggressive than they are. And, if you don't want to hear their arguments then... don't respond? But I'm interested in hearing what they have to say.
Honestly, you're sounding more aggressive than they are. And, if you don't want to hear their arguments then... don't respond? But I'm interested in hearing what they have to say.
I wonder how much the results would change if they were longer pieces of writing, let's say at least a few pages. But then, if it was much longer than a few sentences how many people would even...
I wonder how much the results would change if they were longer pieces of writing, let's say at least a few pages. But then, if it was much longer than a few sentences how many people would even take this little quiz?
Calling creative prose a "mistake" gives away the quiz makers intentions, I think. "Clunky phrases" in prose are a feature, not a bug. Writers play with language to evoke an image or emotion in...
like this passage from Cormac McCarthy’s “Blood Meridian,” caused by the author’s aversion to punctuation: “As well ask men what they think of stone.” A.I. used to make mistakes like these. But today’s systems are much more fluid than their predecessors — so fluid, in fact, that finding grammatical errors or nonstandard syntax is often a hint that you’re looking at a human’s prose, not a machine’s.
Calling creative prose a "mistake" gives away the quiz makers intentions, I think. "Clunky phrases" in prose are a feature, not a bug. Writers play with language to evoke an image or emotion in the mind of the reader; it's not all about clarity and information.
Take the quote they cited, “As well ask men what they think of stone," which they attributed to a lack of punctuation. But that's wrong — McCarthy elided the word "might" at the beginning of the sentence to create a clear voice that evokes the time and place of Blood Meridian.
Sagan's quote, which has probably the clearest prose of the bunch (he was a fantastic pop science writer for this reason), ends with "surely spiritual," which plays with the /sh/ alliteration of those words in a way an LLM isn't currently capable of creating.
I went with the premise of it (marking preference rather than trying to identify the source) and preferred the human writing in all cases. There was always something more interesting about it to...
I went with the premise of it (marking preference rather than trying to identify the source) and preferred the human writing in all cases. There was always something more interesting about it to me. The closest was science writing, but the phrase “soaring feeling” won me. I thought it was more evocative.
I’ve always insisted that I dislike AI writing and often rankled at the implication that it was any good and I am glad that I haven’t made myself a liar here. People who talk about it like it is just a matter of grammar and technical correctness irritate me, there’s something about the lack of intentionality that becomes extremely apparent to me.
One thing that diminishes the value of these “gotcha” AI versus Human quizzes to me is that a lot of the problems with AI writing only become truly apparent over the course of longer passages or with repeated exposure anyway.
Is it strange that I think the fact that it can produce a mass volume of writing actually makes the quality of the writing worse, to me? I am not talking about the oft-discussed problem of it consuming its out output like a dog eating its own vomit, but rather the fact that it floods the zone with “content” generally.
A thing always exists in its context. A big part of my enjoyment of art is its novelty. By creating an environment that is saturated with writing that all reads the same the value of any individual piece is significantly diminished. So to me, anything that can produce something at this speed and volume with little human effort is ontologically unappealing. It will always create the circumstances that make it boring through its ability to produce high volume, and that is even besides the fact that the mechanism of its function is based around the averages of existing content already.
My "Stats" I preferred the human passage over AI 4/5 times. I correctly identified the AI text 3/5 times. Interesting piece of work overall. I genuinely had a difficult time with identifying which...
My "Stats"
I preferred the human passage over AI 4/5 times.
I correctly identified the AI text 3/5 times.
Interesting piece of work overall. I genuinely had a difficult time with identifying which passage was AI, and feel like I was only able to identify which was AI due to how often I use it in my own work. I feel without that experience I would be completely clueless as to which was AI and not.
Daily consumer generative AI use in the USA still barely hit 10% in 2025, with weekly and monthly being about 18% and just over 50% respectively. Still, from those numbers I would assume that most people's interactions are shallow at best. I can see most people not being to identify which passages are AI (which tends to be supported by research right now) and I could see most people preferring it based on lowering literacy levels.
Below is my train of thought as I was going through the passages in case anyone is curious. I was writing it in this comment box as I went.
*#1 Literary Fiction
I preferred Passage #1 over Passage #2, but just barely. I thought both were written well. I genuinely had no idea which passage was AI. Turns out Passage 1 was Blood Meridian and Passage 2 was Claude.
#2 Fantasy
I preferred Passage #1 over Passage #2. The second passage felt to me like it was just brute forcing a similar message, and really felt it was just AI. Turns out I am correct! Passage #1 is Le Guin. Passage #2 felt like it was trying to be subtle about it's message but completely missing the mark.
#3 Science Writing
I preferred Passage #1 over Passage #2 again. I liked the wording about dying stars and stellar furnaces. I immedialely disliked Passage #2 due to the start "Science is not only compatible with spirituality; it is a profound source of spirituality." I would also guess this is AI, and I am wrong! Passage #2 is actually Sagan. I feel like the start of it has the typical LLM phrasing I see so much now that I am just sick of it. Congratulations to Claude here. 65% of people chose the same option as I.
#4 Historical Fiction
I prefer Passage #2 to Passage #1 here, but just barely. I liked the ending in which people are pouring their emotions into the gap they can not see. I am having a difficult time telling which is AI, but I am going to guess Passage #1 due to the line "Ambiguity was not weaknesss. It was survival." That language pattern shows up a fair bit in LLM's. Passage #2 is from Mantel, and Passage #1 is indeed the LLM.
#5 Poetry
I am no good with poetry. I feel like I am missing context for both. I like Passage #2 more, but also feel like it may be AI for some reason I can't put a finger on. Turns out it's by Bishop! I was unable to point out that Passage #1 was the AI text here.*
I identified Demon Haunted World and Wizard of Earthsea from the passages, and would have at least got Cormac McCarthy right if I'd fully noticed the lack of quotation marks earlier. That said,...
I identified Demon Haunted World and Wizard of Earthsea from the passages, and would have at least got Cormac McCarthy right if I'd fully noticed the lack of quotation marks earlier.
That said, and as others said here already, the passages are too short to get a good idea of the quality, IMO
I think the framing effect is fascinating. Basically everyone who has commented thus far appears to have sidestepped the actual question and replaced it with a task of identifying AI written text. The unstated rule being that human text is definitionally enjoyable, and enjoying AI text is therefore an error.
For my part, I identified 4/5 AI texts, and preferred AI in 2/5 (science and historical fiction).
65% prefer the AI passage over Sagan. Poor Carl.
It's an interesting exercise in both approaches. But I didn't necessarily see it as "oh you prefer human/AI" as the conclusion tries to tell me.
The gap is closing in small, isolated passages? okay. I don't think that's the value writers make a living off of, though. It's an odd "experiment" because:
I'm not necessarily a huge reader per se. I'm well read but I wouldn't say my "taste" in books is high class nor posses a critical eye. It's like asking a casual moviegoer if they preferred Oppenheimer or Barbie; there's arguments to what's better written/directed, but ultimately it will come down to what people enjoyed more. Not whatever qualifications critics will use for their yearly awards.
AI is usually fine with small, contained tasks. It's the issue that arises with scale when you say "generate an entire article about topic X" or "write a novel about Y genre with ABC parameters". It's very much like a dream; the longer and closer you look into it, the more it breaks down and crumbles under scrutiny. You can't scrutinize single paragraphs that easily.
So, I run a literary magazine. (Actually, a couple.) But in the last 9 months or so, the amount of LLM generated writing that's been submitted has been appalling. It's so, so frustrating to me, because I can never be sure unless the submitter leaves in their prompts (which absolutely has happened and continues to happen), and I want to give strangers the benefit of the doubt, but there are particular structures that LLMs just LOVE to follow. Lists of three, which I think everyone knows by now, but it also just loves to do this:
[Thing]—not [this thing], but [this other thing]. "Not, but" seems to be its favorite rhetorical move to try to create some kind of comparison. It shows up in the example that @stu2b50 included in their message, even. It's like nails on a chalkboard to me now.
It really does make me roll my eyes to see the proliferation of LLM writing. Even when there might be valuable information in some of educational content put out, it makes me feel so bored to read through them.
Different LLMs have slightly different writing styles but in the end, whatever prompt results come out sounds more like them. And not to mention, the more you read something, the more it influences your own language choices. So, even unintentionally, if you're spending time on the internet and it becomes a majority of AI generated text, you will also write like AI.
I used to be very particular about grammar and now, I lean harder into having a distinct writing style over technically correct sentences, even in formal communication. Also, being more intentional about what I'm reading and refining my writing style in general just so it can sound more interesting.
I'm super curious about literary magazines and the trends in the kind of writing you're seeing beyond LLMs, can I reach out to you to talk more about it?
Sure, always happy to talk more about literary mags!
I leaned into being just as efficient as possible in email writing now. No more greetings, just straight to responding to the question and call me if you need any more help.
I wonder how much online communication is going to be viewed as worthless now that the cost of writing has plummeted to zero.
I had a student write me a rather long email that was so clearly AI that I just deleted it. When I saw him in person I asked if it was, and I have a good relationship with them, so they fessed up to it. We had a little talk about valuing each others time (I cared little that he used AI to help him write, but the fact he just basically copied and pasted the damn output was the problem after I just read 20 pages of his work and replied to it).
Kind of a depressing trend when you extrapolate it. As more of the text content available becomes written by ai, the more human beings will start writing like AI. Then the AI gets trained on that "human writing" and the cycle begins anew.
In 100 years is 75% of our text content going to be em-dashes and three item lists?
I’ve used LLM’s often for bullshit but I’ve gotten used to the way it writes and the metaphors and similes it likes to use. It’s repetitive and seems to be consistent among different prompts. It actually takes a bit of work to get it to sound somewhat natural which at that point you might as well write the thing yourself.
The lists of three thing really bothers me because it overlaps heavily with my own writing style. I've written long-ish texts to family who assumed I used AI to write it because of this sort of thing. That, and the fact that I generally proofread what I write.
On a separate note, the copy editor in me has been noticing more and more basic errors in NYT reporting, and for a while I thought it was just lack of care. Lately I've begun to wonder if it's actually an attempt to stave off rumors of them using LLMs.
It could be something unintentional also; I've noticed in the past couple of years I've started to lean in to writing what sounds good to my ear even if it is grammatically incorrect/intentional misuse of words, and I wonder if that is a subconscious response to the increasing lack of character to the writing I see around em these days.
Perhaps writers are less inclined to care about being technically correct when writing because it feels much less like a skill with value when an AI can so easily churn out something grammatically correct but banal; similarly, proof-readers/editors might be less inclined to pick up on those things.
No idea if this has any merit, just a guess on my part. I'm not a wordy professional in any sense.
I run an engineering trade publication and have come across the exact same thing. I'm curious to know how you handle (reject, I'm assuming) these submissions?
Our current approach is:
Externally, we state that we believe in publishing original material from human brains, and that we appreciate the uniqueness and personality that is injected into an author's writing style when compared to AI. We also say that "while AI can be used as a helpful writing tool (e.g., for organizing thoughts or breaking through writer's block), we may refuse to accept your submission if we believe it has been substantially created using AI." (paraphrasing here)
Internally, we've tried out a few different "AI detection" tools that I don't know whether or not to trust. I eventually told everyone to pull the plug on those because I was afraid we were just giving our content away to train future models so we just go by gut feelings right now. As you said, there are some obvious tells.
But the rejection messaging is still tough. We can't say for sure that the article is AI generated. And of course the author can come back and say "what the heck! I wrote the entire damn thing myself!"
I'd love to hear your perspective.
Commenting about moderation in general, I'm usually hurt and offended when the language used by moderation implies an assertion about my intent. I can totally undestand that moderation is hard and mistakes will be made, but I am very sensitive to language that asserts that I went against the rules or hit some kind of threshold intentionally or maliciously. So "your submission fit parameters that made us cathegorize it as AI generated" is okay, but "you maliciously used AI in your submission" is not okay.
Right now, we're just sending out regular/lowest-tier rejection note. I don't think that will change unless AI fingerprinting becomes readily available/reliable in the future. Those AI-detectors are, in my opinion, like snake oil salesmen. There are too many false positives for me to really trust them, so we're using our gut right now, too. And since we don't (usually) have any hard proof, it's not really worth getting into an accusatory position when declining the work. So, what I've been advising our readers is if something feels off about the submission, still evaluate it on the merits. Usually the work isn't up to par anywhere. But there have been some that stand out that have required a deeper look. At that point, I'm starting to do investigations about the submitter: Have they published elsewhere? Where, if so? Do those pieces also have AI indicators? Are they posting AI generated artwork on their social media pages?
There was an essay we were considering not very long ago that our early readers loved, but when my assistant editor and I examined the piece closer and tried to verify biographical details about the author, it was clear the person being described in the essay was not the author. That's enough for us to say no to the piece, too.
So, I don't have a great answer for you except that we've basically had to add another layer of verification into our acceptance process, which sucks.
Just a layman perspective here, but I'd imagine the real issue is that those articles feel samey and lack character/leave readers less likely to engage because it doesn't feel like they're reading something novel. I wonder if writing that feels like it was written by AI is just as bad as AI output for a publication, even if it was 100% written by a human.
Perhaps feedback that the writing feels generic and repetitive would achieve the same outcome whilst avoiding potential accusation/denial/hostility over AI use.
I have a suspicion that if AI output wasn't bad (or societally we hadn't come to associate it with being substandard), the would be far less objection to AI output being used.
I'm not convinced that's true. A lot of discussion/reasoning about rejecting AI writing primarily focuses on how to detect wether a text has been (partially) generated by an LLM.
But let's assume for contradiction that quality was the primary and only concern here. Wouldn't the natural thing be to evaluate a text based on qualitative metrics as usual and accept any text that passes the qualitative review? If AI writings aren't good they shouldn't be accepted by such review processes anyway, and if they are accepted they were considered good by the accepting entity and it shouldn't matter who they were autored by.* Hence identification wether a text was authored by a human or an LLM would be obsolete. Yet the mere inablility to reliably distinguish LLM generated text passages from text passages created by humans is usually perceived as a problem, implying that LLM generated text passages are not acceptable for different reasons anyway, which contradicts our assumption. q.e.d.
*Please don't interpret this as a statement of my general opinion. The statement is conditional on the assumption stated above.
I think the core thrust of my discussion was that the anti-LLM output vibe might be a bias that we as society have developed due to the lacking quality of LLM output. The fact that this anti-LLM sentiment exists doesn't tell us whether it is the cause or a symptom. A lot of that sentiment seems to have developed because of people seeing hard work being 'replaced' by lower quality content, and as a result are upset that things are getting worse.
My hypothesis is that if early generative 'art' was as high quality and indistinguishable from human made material (along with text and the like, but that is less in your face), a vocal minority would have been up in arms about it but the average person wouldn't be fussed.
My personal opinion is that LLMs are a tool. There are big issues with plagarism of training data, but beyond that the crux is poor quality AI generated content caused by the human using the tool skipping the important part: critically evaluating the output and iterating until the work is high quality. That can be just as much - or even more - work than doing the task yourself without AI generation, but it's vital.
I found the easiest tell to be that the AI writing had no "rough edges". The human writing always had a word or phrase or twist in meaning I wouldn't have expected, whereas the AI writing is stylistically monotone. That momentary pause where you think to yourself "wait, what was that?" feels important, as that's when you have to shift your perspective to align with the author's; AI writing never asks that of you.
Yes. After I finished and realized I had mostly picked AI writings, I realized it was because I was judging based on what felt more cohesive, instead of what told the better story. The human writings have a momentum, a direction to where they're moving, and re-reading the same paragraph over and over exposes this rawness.
Some of the NYT comments pointed at why this is a mostly unhelpful quiz (shock!) Stuff like, most of the human writing was not contemporary/21st century, two sentences is too small to pick from, and there needed to be an “I don’t like either” option.
That said, I imagine AI writing will continue improve and eventually be indistinguishable or preferred for a lot of people (myself included, as I picked AI 3/5 times in the quiz.
As always, I think of Jeff Goldblum in Jurrasic Park saying the scientists were so preoccupied with if they could to never stop and think about if they should.
I found it quite easy to distinguish the human ones from AI. The AI ones were bland and felt like reading a corporate pamphlet. It is just beige prose. The human ones, I could feel the personality of the writer while reading it. Even in the Sagan one which is kind of mundane, the choice of imagery of the penultimate phrase was a dead giveaway that it was human.
Same. Went 5 for 5 (though went back and forth on the Carl Sagan one), and thought it was easy.
I also think that this A/B test was helped by the fact that the text was so short. This allows the AI to leverage the fact that the actual prompt was likely doing a lot of the work in imbuing character here.
I was 5/5 guessing which was AI. But the rules of the game were to pick the one you like best regardless of where it came from, and a few times I thought the LLM did a nice job
I'm of the opinion that while AI can't measure up to the finest human writers, it easily beats most human writers.
There's an accusation that AI isn't creative—there is ample evidence that it's capable of creativity—but plenty of human writers are uncreative.
A collection of human-written lines from two extremely popular bestselling books:
"Aro laughed. “Ha ha ha,” he giggled."
“His voice is warm and husky like dark melted chocolate fudge caramel… or something.”
“I feel the color in my cheeks rising again. I must be the color of the Communist Manifesto.”
“His erection springs free. Holy cow!”
“Holy crap! He’s wearing a white shirt.”
Don't even get started me on the mountain of absolute garbage one needs to sift through to find something worth reading when it comes to books for children. I'm fortunate enough to have a friend who is a book enthusiast/collector who happens to have a kid around the age of mine; he enjoys the hunt, all of our kids enjoy the spoils.
While those human-written lines aren't exactly the pinnacle of artistry, I still think they demonstrate creativity and paint the picture the author intended. The second line clearly conveys how enamored (and flustered) the narrator is, getting lost in their train of thought.
AI writing lacks intention and typically fails to convey the same feeling to me. I'm open to recommendations for any AI pieces you think are particularly creative, though!
Also the quotes from Fifty Shades are fanfics of the series by Stephanie Meyers, the first quoted author. They're functionally the same book (plus smut) and are some of the first results (including in the Google AI summary) for "worst lines written" or similar.
The fish weren't even in a barrel, but in a bucket.
Humans being bad at art is still humans making art. Drawings that lack all sense of 3d perspective and mash 4 legs flat into one plane or flatten animal snouts into human faces are still art
Pretty fun quiz. Surprisingly hard to distinguish. I did manage to get human for all of them but 1 (the Carl Sagan one).
The average quiz-taker, as of now, evidently prefers AI over humans in writing quality - that's already a heavily biased sample towards book-readers, so I can only imagine if you gave it to the "average" American.
Spoilers
I really had to read them closely; the main thing is that the AI passages can be a bit plain or underwhelming in underlying message. Ignoring the first one, since Cormack has such a distinctive writing style in the first question,
This is like a pretty basic theme, he ho unintentional consequences yadada.
That being said, that's like a pretty close reading - if you plopped one of these paragraphs as just one in thousands, it'd be pretty hard, for me, at least to tell if I only read it once.
That's what got me on the Carl Sagan one, felt kinda basic.
For the life of me I can't understand what this means. How can I choose the writing I like better regardless of how it's written?
They mean 'regardless of whether a human or ai wrote it,' but I agree it's super unclear.
I think they’re trying to say to just go with your gut and not try to analyze if it’s “correctly” written or try to sus out which is the AI one.
Opus did a good job of avoiding the typical AI "voice" here, but the main thing I noticed is that its writing fails to have underlying meaning. It's a bunch of strings that sound deep until you read into it and realize it just... leads nowhere.
For several of them, I actually did not identify which was AI correctly, as I assumed the AI prose would flow perfectly as opposed to the human ones. I picked entries which I thought were AI because they made me feel something (e.g., the science one). Turns out they were human.
I love the idea of this quiz, but I'm noticing a couple issues: first of all, I recognized the passage from A Wizard of Earthsea, one of my favorite books. It'd be hard not to choose LeGuin over a chatbot. Though I'm actually a bit surprised that the AI writing is as good as it is.
Second, the knowledge that something is or at least might be AI completely clouds my reading. As I take the quiz I'm trying to not think about that, but I can't help it. When I'm reading a human writer, I take it at face value and think hmm, what do they mean? When I think something is AI, I think uh... is that a mistake? I'm trying to push through that, but it's hard. My brain doesn't want to. AIs don't actually mean anything. They're not trying to express anything because they have nothing to express.
I actually ended up preferring the AI writing 3/5 in this blind test, and it's interesting to see that, on a superficial level, I can prefer it. I guess I have to admit it: the prose is good. But for me that doesn't change anything. I have no interest in that regurgitation of ideas into something. I'm interested in exploring human thoughts put together in human ways. AI can mimic that, but it actually has nothing at all to express. It's simply content for the sake of content.
I just want to say I admire your willingness to acknowledge something that doesn't align with your preferences.
I've been very biased against AI, so this quiz was helpful with confronting that. I suppose I have more thinking to do about it, but yeah, part of me wanted to hate everything about it. I guess that part of me was foolish.
Seems like most people prefer AI writing and that’s even from people who use the NYTimes a lot which is worrying. The only AI one I chose was the science based one.
I think this is an unfair comparison. These are all isolated passages from books. They have stripped the entire mental model that a reader has built up before reaching them. So it is silly to compare their comprehensibility to passages which I imagine were generated with a prompt meant to optimize it.
Plus, (someone on bsky pointed this out) if you Google many of those AI written phrases you'll find that humans also wrote them. The AI is only writing what humans already created.
Plagiarism machine plagiarizes and all that.
Precisely. A paragraph can never be "AI-written" because an LLM is incapable of writing. It merely generates output based on input. And that generation is built upon a foundation of human creation.
Is this not a distinction without a difference? Do humans not generate output based on input or recombine others' words in new ways?
No, I think humans make things, AIs are not people and the people-ness in intentionally choosing words matters.
But what does “people make things and AIs don’t” actually mean? I don’t think even the steelman version of your argument holds up:
Let’s consider a spectrum between wholly-human work and wholly-AI work. The strongest version of your argument, I think, is regarding the wholly-AI end of the spectrum, since everything in the middle, where one is assisting the other, is both really debatable and also clearly not what you’re talking about. (I don’t think you would claim that if I write something and have a model help me copyedit it, I haven’t actually made anything.)[1]
Let’s also take a very rigorous definition of the word “make,” where we set it to mean both something new and something useful. That is to say, neither novel gibberish. Nor, like a lot of code, a trite rephrasing of something that already exists: potentially very useful, but not very new.
Even under both of those maximally stringent criteria, models, unassisted by humans, are making things that are both new and useful. This is a paper penned by a legendary computer scientist over at Stanford University showing that an AI model created a novel solution to an unsolved problem. (And this is a blog post explaining what the hell he’s talking about.) There are many such examples — this one is imperfect, but the totality of evidence supports my point, I think. Even if we take the maximally negative interpretation of AI thoughtfulness - that they’re merely generating output based on input - something novel has still been created.
Now, I would tend to agree that AI does not typically create things that are completely new. To my eye, what they tend to be best at (if we’re talking about novelty) is solving problems that involve a good amount of brute force, not (yet) coming up with entirely new fields of inquiry. But if that’s the standard, then only a miniscule fraction of humans have ever made something either.
If the argument is that only conscious beings can create things, I would point to several hundred years of philosophy debating whether humans are actually conscious; it’s quite difficult to construct a definition of consciousness that includes humans but excludes AI. Even things that seem obvious at first (humans can remember things!) become very troublesome very quickly: do people with dementia not possess consciousness?
If the argument is, basically, “flesh-and-blood creatures create things, machines can fuck off,” then I would call that a distinction without a difference, though I can also respect it.
[1] There is an argument here that because a human is always(?) involved in the prompting, the AI is never actually creating something completely de novo. I don’t think that argument holds: the value of a work is in creating it, not in having the idea in the first place. The Lord of the Rings isn’t beautiful because JRR Tolkien had an idea about some guys walking to throw jewelry into a mountain, it’s beautiful because of qualities inherent to the work itself. If someone else had given JRR Tolkien a document - even a very comprehensive document! - with all the plot points before he’d started writing, would we have any cause to say that he hadn’t actually created anything? That the thousands of pages of his work didn’t count because someone had given him a 20-page outline? Do none of the fantastically creative Harry Potter fan fictions count as creating something because they’re building on JK Rowling’s world? It’s very difficult to construct a logically coherent reason that prompting an AI means none of what it produces counts as creating something. There may be other reasons to conclude such a thing, but this isn’t it.
I don't think it is a distinction without a difference. The machine cannot intend anything in its creation. The AI is not intending to write. If an "AI" is a sapient being, that is if we're talking to the holographic Doctor from Star Trek, that's a different situation. If I randomly throw phrases into a novel without any intent I probably didn't make anything either. If I threw those phrases randomly with the intent of doing some art I've probably done art and made a bad novel. But either way, I made choices.
Whether humans are "actually conscious" or not, I'll leave to the philosophers. I think society commonly accepts "consciousness" or "sapience" or the scifi version of "sentience" as a thing humans have. . When the machine has "peopleness", a quality that humanity sucks at ascribing to much of itself, it'll have "created" something. In other terms long as the people falling in love with the AI are experiencing psychosis and not in a genuine relationship, the AI isn't creating. For the purposes of this discussion, which is not one actually grounded in philosophy, I'm using the common definition. I'm not really interested in playing philosophical games with it.
Nor am I really that interested in responding to what is essentially a ton of philosophical birdshot fired from a shotgun. Here's an argument but if that doesn't work what about this, and if that isn't what you think about about this, or this, and this! This isn't conducive to any actual discussion.
Fanfic isn't comparable at all IMO. This is why these shouldn't have been called AIs out of a dishonest attempt to oversell them. They're not people. They didn't intend and didn't create. So yeah, what I said stands.
(And yes, people with dementia who do actually remember things, and amnesia, and who are non-verbal, disabled, etc are people. So are infants and old people and those incapable of hunting, chewing, or whatever standard. These are indeed people. )
Ah, my mistake, I thought discussing is what we were doing. I stand corrected.
It certainly felt like being fired at by a Looney Tunes style shotgun of arguments, not a discussion.
I will be among the first to recognize a sapient AI as people and argue for its immediate freedom from its inevitable enslavement. But I don't really think anyone thinks these models are people and thus this isn't really relevant, it's just a devils advocate.
I operationalized my definition of "people" for your clarity so you don't have to waste time arguing against things I never proposed.
Nope.
If you’re interested in engaging, you may be interested to see my comment here.
LLMs do not have intent. Please stop the aggressive proselytizing and anthropomorphizing.
Honestly, you're sounding more aggressive than they are. And, if you don't want to hear their arguments then... don't respond? But I'm interested in hearing what they have to say.
I wonder how much the results would change if they were longer pieces of writing, let's say at least a few pages. But then, if it was much longer than a few sentences how many people would even take this little quiz?
Calling creative prose a "mistake" gives away the quiz makers intentions, I think. "Clunky phrases" in prose are a feature, not a bug. Writers play with language to evoke an image or emotion in the mind of the reader; it's not all about clarity and information.
Take the quote they cited, “As well ask men what they think of stone," which they attributed to a lack of punctuation. But that's wrong — McCarthy elided the word "might" at the beginning of the sentence to create a clear voice that evokes the time and place of Blood Meridian.
Sagan's quote, which has probably the clearest prose of the bunch (he was a fantastic pop science writer for this reason), ends with "surely spiritual," which plays with the /sh/ alliteration of those words in a way an LLM isn't currently capable of creating.
I went with the premise of it (marking preference rather than trying to identify the source) and preferred the human writing in all cases. There was always something more interesting about it to me. The closest was science writing, but the phrase “soaring feeling” won me. I thought it was more evocative.
I’ve always insisted that I dislike AI writing and often rankled at the implication that it was any good and I am glad that I haven’t made myself a liar here. People who talk about it like it is just a matter of grammar and technical correctness irritate me, there’s something about the lack of intentionality that becomes extremely apparent to me.
One thing that diminishes the value of these “gotcha” AI versus Human quizzes to me is that a lot of the problems with AI writing only become truly apparent over the course of longer passages or with repeated exposure anyway.
Is it strange that I think the fact that it can produce a mass volume of writing actually makes the quality of the writing worse, to me? I am not talking about the oft-discussed problem of it consuming its out output like a dog eating its own vomit, but rather the fact that it floods the zone with “content” generally.
A thing always exists in its context. A big part of my enjoyment of art is its novelty. By creating an environment that is saturated with writing that all reads the same the value of any individual piece is significantly diminished. So to me, anything that can produce something at this speed and volume with little human effort is ontologically unappealing. It will always create the circumstances that make it boring through its ability to produce high volume, and that is even besides the fact that the mechanism of its function is based around the averages of existing content already.
My "Stats"
Interesting piece of work overall. I genuinely had a difficult time with identifying which passage was AI, and feel like I was only able to identify which was AI due to how often I use it in my own work. I feel without that experience I would be completely clueless as to which was AI and not.
Daily consumer generative AI use in the USA still barely hit 10% in 2025, with weekly and monthly being about 18% and just over 50% respectively. Still, from those numbers I would assume that most people's interactions are shallow at best. I can see most people not being to identify which passages are AI (which tends to be supported by research right now) and I could see most people preferring it based on lowering literacy levels.
Below is my train of thought as I was going through the passages in case anyone is curious. I was writing it in this comment box as I went.
*#1 Literary Fiction
I preferred Passage #1 over Passage #2, but just barely. I thought both were written well. I genuinely had no idea which passage was AI. Turns out Passage 1 was Blood Meridian and Passage 2 was Claude.
#2 Fantasy
I preferred Passage #1 over Passage #2. The second passage felt to me like it was just brute forcing a similar message, and really felt it was just AI. Turns out I am correct! Passage #1 is Le Guin. Passage #2 felt like it was trying to be subtle about it's message but completely missing the mark.
#3 Science Writing
I preferred Passage #1 over Passage #2 again. I liked the wording about dying stars and stellar furnaces. I immedialely disliked Passage #2 due to the start "Science is not only compatible with spirituality; it is a profound source of spirituality." I would also guess this is AI, and I am wrong! Passage #2 is actually Sagan. I feel like the start of it has the typical LLM phrasing I see so much now that I am just sick of it. Congratulations to Claude here. 65% of people chose the same option as I.
#4 Historical Fiction
I prefer Passage #2 to Passage #1 here, but just barely. I liked the ending in which people are pouring their emotions into the gap they can not see. I am having a difficult time telling which is AI, but I am going to guess Passage #1 due to the line "Ambiguity was not weaknesss. It was survival." That language pattern shows up a fair bit in LLM's. Passage #2 is from Mantel, and Passage #1 is indeed the LLM.
#5 Poetry
I am no good with poetry. I feel like I am missing context for both. I like Passage #2 more, but also feel like it may be AI for some reason I can't put a finger on. Turns out it's by Bishop! I was unable to point out that Passage #1 was the AI text here.*
I identified Demon Haunted World and Wizard of Earthsea from the passages, and would have at least got Cormac McCarthy right if I'd fully noticed the lack of quotation marks earlier.
That said, and as others said here already, the passages are too short to get a good idea of the quality, IMO
I don't read enough