I’m in both a lot of art based communities and programming based communities (like HN). The one blind spot many HN readers have is that the quality of something is one facet of an artwork.
Humans make minute decisions in their art whether to tell a story or represent something. When someone generates something, at some point you as the viewer/reader could assume that arbitrary choices were made in places. Those arbitrary decisions just aren’t interesting. I’ve quoted this before, but Kane Parsons highlights exactly this thing:
> If I see an element of the environment has been, you know they used generative fill or whatever to change something about the scene, even if--. It just shuts the part of my brain that wants to know more about that world and wants to look for details, 'cause I would assume if they're willing to make an arbitrary choice there, they can make an arbitrary choice with literally anything.
It’s just less interesting to our minds, I think because there’s less to ponder and explore. It can be a masterpiece in quality but when I know something in that piece was just made up by a machine, why would I care? I can’t wonder why the artist would include it. I can’t think about what it’s trying to tell me because it’s likely not telling me anything.
YES. Well said. This is why I want to know which text is from human thought labour. To read text or to consume outputs of human minds is to learn about those minds and how those minds might work.
There's a deep loss in having no mind to learn about in the subtext of the specific knowledge content written... I never knew this potential was there (for me) until it started to disappear as a domain of my interest and enthusiasm
I'm definitely a collectivist. Maybe this is related to that value system, or some other subtle thing that I didn't know varied in experience until a fracture and unbundling started. It feels a like how COVID introduced new fracture points in worldview and value systems that became less tenable to ignore, or at least harder to bridge and work across
It also just occurred to me, inspired by your comment, that getting to know a mind is more than just satisfying your curiosity; personal experience confirms for me that as soon as I interacted with a brilliant mind, some of that shone back on me, and made me a tad more brilliant as well.
We always want to be among the people we imagine or visualise our future better selves to be. It's such an absolutely fundamental part of humanity (it is how children become adults, too)
The reason "AI artists" engage in a spiral of angrily and defensively explaining that their process is not just pushing a button, or trying to project AI-like attributes onto existing things like "covering someone else's song" (which is what the Fender CEO recently did!!) is that we all know that, deep down, using AI to do creative things people did is less in at least some way. It's avoiding some emotional labour.
People who know this less is a big part of what they do are the ones who need to reckon with it.
Everyone else is free to mentally downgrade it just like they also mentally downgrade excessive photoshop, overproduction, etc., and they will.
It took a generation for the world to find any kind of accommodation with what photoshop is capable of, what is appropriate and what is not, and there are aspects we are still justifiably angry about (fraud, deception, body image). AI artists seem in a hurry to have their own use normalised; they want people to hurry up and get over it and use it themselves so they don't feel bad.
Many photographers have developed rules for what they will and will not retouch, even a century before photoshop — for example, if the spot on someone's face won't be there in a fortnight, it maybe even shouldn't be in a photograph that is intended to be a long-lasting, somewhat authoritative representation of that person. But this took a long time to become any kind of cultural norm.
I think AI-only coders and CEOs are also in this "hurry up and get over it so I feel OK about my choices" phase, and I think we need to see this in a clear-minded way if we are to confront its impacts.
I think it’s a blind spot to a lot of people not just programmers. If you don’t have experience understanding and not just viewing you’ll almost always prefer the shinier object. I don’t really think this is particularly new even generative AI has just ramped up the speed that shiny objects can be produced. Most people aren't trained to do the understanding part so most media has always been about maximizing the shininess.
I don't think reading isolated stories holds much value. The value of the emotion it elicits is much lower than knowing the person who wrote it or in what historical or social background they did so.
Reading stories written by humans is one of the ways we can keep ourselves grounded to other humans: be it through fantasy, lyrics, or autobiographies. I can only describe reading AI stories as emotional indulgence.
> The value of the emotion it elicits is much lower than knowing the person who wrote it or in what historical or social background they did so.
Then the value of many ancient texts must be close to zero, because we often know nothing about the people who wrote them. In many cases even our guesses for where and when they were written differ by multiple centuries or thousands of kilometers.
No that makes them more interesting for the same reasons OP gave. You try to build a mental model of this mystery person from a mysterious time, where ever little detail could be a clue about them, their society/culture, or just trivial.
With AI writing the answer is that everything is randomly averaged and thus trivial.
> You try to build a mental model of this mystery person from a mysterious time
I’ve read many ancient texts with great delight, and not once have I tried to “build a mental model” of their authors. It’s a dimension of literary analysis that has never interested me, even for authors (such as Kafka) whose personal lives are generally considered inseparable from their works.
The work stands on its own for me… or doesn’t, in which case learning about the circumstances of whoever wrote it won’t save it in my eyes.
Every story told, is folded into the corpus of all human stories, all variations of the heroes journey, all variations of the hammon-cycle. AI by being trained on the corpus, is average - which makes it better than 80 % of mankind.
But we are tired of many stories. We see "new" movies until we are around 20 - then the repetition sets in- we get a refined taste and long for the exotic (-30) until finally, we search for outliers, for anything new, desperate under the sun.
AI can not provide that. But also humans can not provide that. Every bard, every writer, every poet, every band rises every morning to compete against the beatles. Most of them loose and thus abandon the craft to win bread elsewhere.
> But also humans can not provide that. Every bard, every writer, every poet, every band rises every morning to compete against the beatles. Most of them loose and thus abandon the craft to win bread elsewhere.
Most, but not all. For every 99 who give up and go and do something else, one transcends.
AI cannot create. Humans can. Most do not, but the possibility is always there that they could.
Not something novel, usually - but i can reproduce recombination results with intuition, which is kind of close enough. Still it lacks the striking direction of an artist/great mind and averages out.
but most of what humans create isn’t novel, esp in saturated areas like art. so even LLMs can “just do that” they’re very capable at outputting worthwhile materisl
If you seriously dig into literature or film you'll find that many of the stories don't fit into the hero's journey at all--or only fit if you make the "journey" so generic as to be practically tautological.
As for the claim about people's lives. People who are serious about art almost invariably keep enjoying the "basic" classics; many musicians keep listening to Bach all their lives. Faulkner claimed to read Don Quixote over again every year. It is possible to find novelty in the well-worn.
Here's a scoop from inside writing circles (pre-AI era): we were a lot into "good writing techniques". Not into communication. We wanted to produce a "marketable product". On the way, we (writers, literature professors, critics, etc) trained readers and writers to appreciate technique by telling them there was such a thing. We established tropes, because "This enemies-to-lovers story is good" sells more than "you should read Book". AI is very good at techniques and tropes. If readers want authentic human writing, they will have to change.
It's quite likely that many survey participants were AI bots, and hence preferring AI generated content. And certaintly, that of the 1600+ people whose 15min they rented, few were paying any meaningful levels of attention to earn their 3 USD.
I cannot judge fairly given that the stories are labeled, but the story about the koi is dreadful. It is about a supposedly mysterious subject—the machine at least insists it's mysterious—but it is so dull, so pedestrian, so cut-and-dry. "My mommy dying makes me sad but the peaceful koi pond helps me find peace". Blech.
Perhaps it only seems so banal because the competition is so engigmatic.
And what does it mean for something to smell of "sterile detachment"?
Yes! Reading all samples really drives home the study's conclusion that “people generally prefer predictability, because difficult material requires more brain power.” I could blow through the AI generated versions as I do Brandon Sanderson or JK Rowling because my brain was saying "yada yada yada, yeah I get it, how many sentences do you need to say the same thing."
At first I was impressed by the “generated” stories and wondered about the prompt used to get such human sounding writing, except then I read further and saw they put the human written stories first and the AI stories after, which are all full of the familiar stylistic constructs (and banality) of AI written works.
Honestly much more curious to know about the human written stories were here! Feels at least pretty important to the study. With the AI, the specifics feel less important, because the relative input is going to be so small, so the researchers can probably just do it themselves. But how do you choose which works to stand in as pieces of human writing in general? How did they make sure the subjects weren't aware of the authors they chose?
"We assume creative writing requires uniquely human qualities, such as emotional understanding and lived experience, which leads people to underestimate AI's capabilities."
It's not about the relative "creativity" of it. Art is fundamentally about communication. It's a social activity, whether you're reading a book or listening to music or looking at a painting. Even a mid-average author is going to make something more compelling, on a human level, than extraordinary prose generated by a machine.
AI-generated creative artifacts are as interesting as the fractals in a snowflake. Interesting, sometimes beautiful, but void of human meaning.
Is art what is generated, or is art what is in the eye of the beholder?
Do we find artistry in nature?
What does it say about us that the sum average of us disgusts us on a visceral artistic level, is that the machine's fault for reflecting us accurately?
Art is a social activity whereby a human uses their creativity to produce an artifact that expresses some aspect of their experience, using non-standard forms of communication. It's a way of signaling to others "do you see what I see? do you feel what I feel?" The beauty of art exists on two layers: the aesthetic quality of the work, and the sense of shared experience among those who consume the art.
We find aesthetic beauty in nature; we do not find artistry.
What was the quality? AI generally will have better grammar and such vs what most people write. So I could believe you prefer the AI content in this example because it is better quality, but this tells us nothing about AI vs human written when someone takes the time to edit their rough drafts.
(Note that hacker news discussions are not conductive to spending days crafting a good response - that takes days - so everything here is a rough draft)
Here are the stories, sadly, they immediately reveal the source so you can't play the quiz really. On the other hand, I think most people would spot the classics like em-dash and metaphores straight away. https://osf.io/d8f6m/files/osfstorage
The real question to be debated is how do you deal with a new kid on the block, who is better than you at [whatever].
Denial seems to be the general human default, it's easy and free, but at best it can only delay the inevitable.
Accept and adapt is probably the most logical approach, but that requires a lot of effort and risk taking, and still will likely leave you net worse off.
I suppose hostility and violence is another human favorite, but that almost never has gone over well.
This sounds as of AI has wider appeal. Like those bland mass-market beers that nobody likes very much but nobody hates either. Restaurants have those on the menu as a second choice.
While a real, published, author has taste and appeals to some people, not to everyone. Journal editors make sure of that.
These are ~1000-word stories (prompts at [1]), squarely in the sweet spot where planning logistics are minimal, context windows comfortably fit everything, and the appropriate level of abstraction omits any details which could glaringly reveal weak world-model knowledge.
SOTA LLMs from 2 years ago would saturate this study, and while modern LLMs are amazing and I suspect they would do quite well on a similar but more challenging study that forces them out of said comfort zone, this isn't that.
No, I have 0 interest in AI stories or music or art. I like all of these things because it's a view into the mind of the artist and I want to connect with another human and appreciate what they made and why.
I feel like this is just "we lied to people and we are surprised they trusted us" which just emphasize how hard we should work to enforce generated content as such (which I don't even know if it is possible in a reliable way)
Anyway, the good thing with human mind is it can explore, where a LLM is statistically producing words. It will certainly produce average stories, which if they are short will be better than most human writing. But it won't explore new ways of writing, which are the ones which can become a masterpiece IMHO
I would be interested in long stories experiments though, is the context wide enough to keep consistency and mid quality or are they becoming boring and un-readable like claude can be when trying to explain a concept ?
Then starts the philosophical questions : If a human need to write world rules, names, locations, characters, psychology, how each character perceive the world despite its rules, define relationship between characters and then delegate the story writing to a LLM, how much of the story belong to the human and what's the difference between doing this and hiring a ghostwriter
If it's not already the case, it likely will soon be that people prefer short main motifs for songs generated by AI over humans.
It could still be quite some time before they're writing entire songs and books, though.
No one buys 1000 word stories, or listens to 12 second songs on Spotify, so for the time being it's not really any cause for concern.
If Facebook has demonstrated anything it's that people have preferred AI memes for a while already.
People "prefer" lower quality art all the time - look at what's big on the radio and in theaters, what's popular on TV, etc.
Why should I be concerned if AI is making trash art instead of humans?
I loved Britney Spears when I was 9. Is the world really worse off if I'm listening to an AI version or Toxic or the Britney Original?
I can see why people would be scared if AI was pumping out Purple Rains and Superstitions on the regular, if it was writing a Secret History every hour, but it's not, nowhere even close.
This just seems like people being grumpy that people prefer Marvel movies over what they think are good movies.
Human artists have been pawning off derivative trash as art for a very very long time, and people have been consuming it and loving it. I'm not going to lose sleep if in 10 years AI is pumping out the derivative trash personalized to the exact level of trashiness desired by the consumer.
No one really knows how this will play out, but it's still uncomfortable to think about robots figuring out how to reliably make things that look good, sound good and are full of meaning, would people still have any motivation to create? Would anything be considered good if everything is "good"? Of course things could turn out fine, I just wouldn't minimize the significance of what's happening.
>It should be noted that these stories were 1000 word stories
A huge limitation of the study IMO.
Most people who read are not reading 1000 word short stories. If they could replicate this story using novella length stories (25-40k words), I'd be curious to see the results. I would wager that a much larger % of people would prefer long form content from humans.
Funny. It's the same result as that famous painting labeled as AI-generated and people found many faults with it. It's just showing that people are biased against AI itself; the quality of the content doesn't really matter.
LLM writing is optimized, at a great expense, for that specific parameter. They literally have human raters fine-tuning "favorability" - general appeal - as a part of post-training. So yes, the conclusion here doesn't surprise me at all. By the way, this is also true of HN! You think you're better than that, but at this point, on any given day, we have around 10 LLM-generated stories on the front page (and another 10 are usually AI-related press releases or op-eds).
Anyway, human writing isn't like that. You have a style that some people will like, some people will find "meh", and some people will hate.
A better question is, do people want to read AI text all day? And the answer to that question is, I think, very different. Most people don't. Which is why so many parties are really determined not to tell you.
One thing I am getting really tired of is the growing inability to distinguish legitimate sources. I don’t want to have to play, “can you tell the difference?” It’s not fun being had by a liar.
I want human written content or nothing at all. I care about the struggle. I like the spelling and grammar mistakes. I enjoy the story between the lines. I like the growth of the author and watching them and their work evolve.
Generated fiction is boring because it has none of that. There is no life in it. No author.
“A human wrote it,” is a light trick. Ok, who? Where are they from? If you have to lie to give answers in order to convince someone they like the story you’re not proving that people prefer AI stories in my estimation: you’re proving that the test is a convincing lie. And maybe that people are gullible.
Stories are always based on realities experienced by the authors, either personally, or viscerally. AI has no experience of this kind.
The experiences then are assembled or re-assembled into an art form, that makes impact on the thinking reader. The key ingredient is how this experience was meaningful, useful, or otherwise impactful enough to have qualified to become art.
Thus, reading fiction (good ones, not the rubbish kind) lets us experience those moments without the actual life-level cost (the only cost is reading). And therefore, these experiences get transferred to us, making us wiser, more aware, more rational-- if the art is real art.
Take the experience-as-art out of the equation and all is left is just slop.
Very bleak. I think like most people, I would prefer for stories not to be written by AI. I value human creativity, and believe that an LLM cannot truly produce art.
This article suggests that when people are lied to that stories were not written by AI (when they actually were), they prefer them.
For practical questions, I always prefer Claude's answer to a human's. Non-narcissistic and to the point. For entertainment, I'm less impatient, although unsurprisingly, most people view reading as homework and are happier to blow through a book written at a YA level. Brandon Sanderson a'int the king for nothing.
"Art is in the eye of the beholder" - it is me, who enjoys art and how it affects me and works with my emotions that is affected by it. Who or what the artist is, should be irrelevant, unless the artist imposes some monetary value I want to profit from at some point.
I don’t think the artist should be irrelevant - the art is the expression of the artist’s emotions. Knowing about the artist unlocks a lot of depth in the art, it’s an attempt to share and express their perspective through creative means.
Not to say how you decide to consume art isn’t valid (there is no proper definition after all). However, you’re doing a disservice to a lot of artists by dismissing their role in it all.
> However, you’re doing a disservice to a lot of artists by dismissing their role in it all.
That is a false assumption. I do not dismiss the artist role, but I also would not be interested in the artist if their art does not move me. People generally do not know who Banksy is, but he is being defined by his art, not by his persona. That may very well happen for an AI agent sometime in the future.
Nonethless, a lot of people enjoy stupid movies, others enjoy smart movies etc.
If you are not like me and you don't care about the time you spend on a 2h movie because you just want to watch something, you will have a complete different experience than me. I watched a 1000 movies and more, I want/I need something 'new' or unique i haven't seen a 1000 times.
Its the same with the AI slop videos; I found the first 3 'cats with jobs ai' videos funny as a category, i don't want to watch thou 'accidents' videos if they are fake.
I'm sorry but all this proves is that a lot of published stories aren't very good. The three human stories attached don't have fantastic plot and read a little poorly.
All three of the human written stories read like they were made for a reviewer/publisher rather than a reader.
I don't think this is a good study or a good representation of good writing. I think many short stories have the issue of force feeding so much information and tricks that the 3 human stories read more gimmicky than the chatgpt ones.
It's pseudoscience no matter what, but I'd like to see the results sliced up for an age range of roughly 35-70. I would heavily question the taste and judgment of those outside that range. Or at the very least note that people's personalities will change significantly when entering/exiting that range.
That's why those "we prefer human writing" polls (such as the one cited in [0]: whooping 98% readers say they prefer human writing to AI!) are just noise.
To quote myself:
> People are really, really good at lying to themselves, let alone to an online poll. They'll tell you that they prefer imperfect or even bad writing as long as it's not AI slop, just like how they'll tell you they like healthier food, they prioritize personality instead of look for potential dates, how they use LLM "only as a spellchecker", and how they use tiktok for educational videos. As long as there is no stake, people will just say what make they feel better.
> That's why those "we prefer human writing" polls (such as the one cited in [0]: whooping 98% readers say they prefer human writing to AI!) are just noise.
Why would they not?
> They'll tell you that they prefer imperfect or even bad writing as long as it's not AI slop,
But I do, and it's very clear that the online communities I frequent prefer people write in their own voice, even with very bad grammar and spelling, rather than resort to chatgpt for their posts and comments — this comes up all the time
> just like how they'll tell you they like healthier food
But lots of us do? Once you are over 30 and you control your own fridge and cook your own food, why would you not?
> they prioritize personality instead of look for potential dates
But (in a fair context for comparison) people actually do, especially women. As long as someone is presentable and shows a bit of self-care and self-respect, for the most part looks rank at least on a par with personality, if not below it.
> how they use LLM "only as a spellchecker"
I don't use it at all for writing; I do know people are weak on this point but I also know that people understand it fosters distrust
> how they use tiktok for educational videos
Huge chunks of people don't use it at all.
Most of what you are saying here sounds like you are projecting low mood and your own personal prejudices and insecurities. Try to be less misanthropic.
Of course more than zero people actually use tiktok for educational purposes and like healthy foods. I'm talking about how questions like that ("do you prefer imperfect human writing or AI writing?", "what do you use tiktok for?", "do you stop reading if you know something is written by LLM?") are ill-designed for questionnaires.
Tests like the OP's link are slightly more meaningful to find out people's true preferences. But of course the best way is still A/B test and tracking user behavior. Anyone has done any form of behavioral tests, online or not, would understand that what people actually do is vastly different from what they think they do. The correlation is often negative.
"Do you prefer imperfect human writing or AI writing?" is a question that is loaded.
It's absolute horseshit (and I would say somewhat sociopathic) to try to A/B test that to catch people out as if to say "aha you are lying to yourself because you can't tell".
People's preferences are representative of their desires and aspirations, their future self-perception, their better self, etc. I'm sad that you seem to think that is something they should be disabused of by statistical means.
It would be misanthropic to assume that because people say they want healthier food but then don't buy it, that they don't really want it. The proper response to that is not to give people what their behaviour says they will actually buy; it is to broadly compel food producers to make healthier food that matches what they wish they would more often choose.
ETA: above all what I want to impress upon you is that "people are really good at lying to themselves" is not actually a statement of fact; it is your opinion and interpretation projected onto the data you choose. There are other ways to explain this gap that do not involve fundamental attribution error.
"Villanova embraces Artificial Intelligence (AI) with a commitment to thoughtful, ethical use rooted in our Augustinian Catholic mission and values. We aim to balance innovation with our humanist tradition and intellectual heritage, preparing students to lead responsibly in a changing world.
Explore resources designed to help you learn, teach and work with AI at Villanova."
Sure studies can be designed to get a desired result, but we're talking about this particular study. Do you see any issues in how it was done that would point to designed-for-desired-result?
Despite post facto conspiracy theories that it was a guerilla marketing campaign to promote "Old Coke", it was not. Coca-Cola tested a new formula in focus groups and people preferred it at the time because, basically, it was sweeter. But this didn't translate into longer term consumption. It's a cautionary tale in how data science can fail and how you shouldn't necessarily extrapolate to the long-term from the short-term.
So when people say they preferred AI written stories, what are they really reacting to? There's probably a lot of elements that are engaging but is it a lasting story? Most enduring stories become so because they speak to some truth of the current times or the human condition that can be hard to separate from the technical merits of the story.
Example #1: Fight Club both as a book and a movie was a bit of a sleeper hit. It was a box office failure. But it became a cult classic because it was one of the earlier works that spoke to the loss of a future for men in our society.
Example #2: The Rocky Horror Picture Show. Again, it was a sleeper hit but it built over time as a relatively early Western expression of queer culture, gender identity and cultural conformance.
Example #3: The Wire. While it had great writing it has endured as a testament to institutional failure at every level and that doesn't necessarily take a villain. Everybody can be acting in their own interests and you end up with inevitable failure anyway.
Example #4: The Great Gatsby. This was never a hit in F. Scott Fizgerald's lifetime. How it became one is an interesting story. Basically, during WW2 the Army created these small print books for deployed soldiers to easily carry and read. Then once the war ended and many of the veterans went to college, they wrote about the Great Gatasby. It's now viewed as an American masterpiece that captures the 1920s like few other works. It expresses class consciousness and some of the metaphors (like the green light on the dock) are debated to this day.
Now, could AI write things that coherently speak to the human condition this way? Not yet (IMHO). Far from it.
That being said, there are only so many base story structures. Take the romance category. You'll find repeated tropes. Could AI write what might best be called "pulp fiction" in that genre? Probably sooner rather than later.
I’m in both a lot of art based communities and programming based communities (like HN). The one blind spot many HN readers have is that the quality of something is one facet of an artwork.
Humans make minute decisions in their art whether to tell a story or represent something. When someone generates something, at some point you as the viewer/reader could assume that arbitrary choices were made in places. Those arbitrary decisions just aren’t interesting. I’ve quoted this before, but Kane Parsons highlights exactly this thing:
> If I see an element of the environment has been, you know they used generative fill or whatever to change something about the scene, even if--. It just shuts the part of my brain that wants to know more about that world and wants to look for details, 'cause I would assume if they're willing to make an arbitrary choice there, they can make an arbitrary choice with literally anything.
It’s just less interesting to our minds, I think because there’s less to ponder and explore. It can be a masterpiece in quality but when I know something in that piece was just made up by a machine, why would I care? I can’t wonder why the artist would include it. I can’t think about what it’s trying to tell me because it’s likely not telling me anything.
YES. Well said. This is why I want to know which text is from human thought labour. To read text or to consume outputs of human minds is to learn about those minds and how those minds might work.
There's a deep loss in having no mind to learn about in the subtext of the specific knowledge content written... I never knew this potential was there (for me) until it started to disappear as a domain of my interest and enthusiasm
I'm definitely a collectivist. Maybe this is related to that value system, or some other subtle thing that I didn't know varied in experience until a fracture and unbundling started. It feels a like how COVID introduced new fracture points in worldview and value systems that became less tenable to ignore, or at least harder to bridge and work across
This.
It also just occurred to me, inspired by your comment, that getting to know a mind is more than just satisfying your curiosity; personal experience confirms for me that as soon as I interacted with a brilliant mind, some of that shone back on me, and made me a tad more brilliant as well.
We always want to be among the people we imagine or visualise our future better selves to be. It's such an absolutely fundamental part of humanity (it is how children become adults, too)
The reason "AI artists" engage in a spiral of angrily and defensively explaining that their process is not just pushing a button, or trying to project AI-like attributes onto existing things like "covering someone else's song" (which is what the Fender CEO recently did!!) is that we all know that, deep down, using AI to do creative things people did is less in at least some way. It's avoiding some emotional labour.
People who know this less is a big part of what they do are the ones who need to reckon with it.
Everyone else is free to mentally downgrade it just like they also mentally downgrade excessive photoshop, overproduction, etc., and they will.
It took a generation for the world to find any kind of accommodation with what photoshop is capable of, what is appropriate and what is not, and there are aspects we are still justifiably angry about (fraud, deception, body image). AI artists seem in a hurry to have their own use normalised; they want people to hurry up and get over it and use it themselves so they don't feel bad.
Many photographers have developed rules for what they will and will not retouch, even a century before photoshop — for example, if the spot on someone's face won't be there in a fortnight, it maybe even shouldn't be in a photograph that is intended to be a long-lasting, somewhat authoritative representation of that person. But this took a long time to become any kind of cultural norm.
I think AI-only coders and CEOs are also in this "hurry up and get over it so I feel OK about my choices" phase, and I think we need to see this in a clear-minded way if we are to confront its impacts.
I think it’s a blind spot to a lot of people not just programmers. If you don’t have experience understanding and not just viewing you’ll almost always prefer the shinier object. I don’t really think this is particularly new even generative AI has just ramped up the speed that shiny objects can be produced. Most people aren't trained to do the understanding part so most media has always been about maximizing the shininess.
I don't think reading isolated stories holds much value. The value of the emotion it elicits is much lower than knowing the person who wrote it or in what historical or social background they did so.
Reading stories written by humans is one of the ways we can keep ourselves grounded to other humans: be it through fantasy, lyrics, or autobiographies. I can only describe reading AI stories as emotional indulgence.
> The value of the emotion it elicits is much lower than knowing the person who wrote it or in what historical or social background they did so.
Then the value of many ancient texts must be close to zero, because we often know nothing about the people who wrote them. In many cases even our guesses for where and when they were written differ by multiple centuries or thousands of kilometers.
No that makes them more interesting for the same reasons OP gave. You try to build a mental model of this mystery person from a mysterious time, where ever little detail could be a clue about them, their society/culture, or just trivial.
With AI writing the answer is that everything is randomly averaged and thus trivial.
> You try to build a mental model of this mystery person from a mysterious time
I’ve read many ancient texts with great delight, and not once have I tried to “build a mental model” of their authors. It’s a dimension of literary analysis that has never interested me, even for authors (such as Kafka) whose personal lives are generally considered inseparable from their works.
The work stands on its own for me… or doesn’t, in which case learning about the circumstances of whoever wrote it won’t save it in my eyes.
> Then the value of many ancient texts must be close to zero, because we often know nothing about the people who wrote them.
This is where imagination fills the gap and makes it interesting to think about.
Every story told, is folded into the corpus of all human stories, all variations of the heroes journey, all variations of the hammon-cycle. AI by being trained on the corpus, is average - which makes it better than 80 % of mankind.
But we are tired of many stories. We see "new" movies until we are around 20 - then the repetition sets in- we get a refined taste and long for the exotic (-30) until finally, we search for outliers, for anything new, desperate under the sun.
AI can not provide that. But also humans can not provide that. Every bard, every writer, every poet, every band rises every morning to compete against the beatles. Most of them loose and thus abandon the craft to win bread elsewhere.
> But also humans can not provide that. Every bard, every writer, every poet, every band rises every morning to compete against the beatles. Most of them loose and thus abandon the craft to win bread elsewhere.
Most, but not all. For every 99 who give up and go and do something else, one transcends.
AI cannot create. Humans can. Most do not, but the possibility is always there that they could.
> AI cannot create
Say what now?
- https://www.jdsupra.com/legalnews/ai-is-inventing-molecules-...
- https://www.forbes.com/sites/anishasircar/2026/04/17/ai-solv...
- https://www.nature.com/articles/s41524-025-01881-2
Not something novel, usually - but i can reproduce recombination results with intuition, which is kind of close enough. Still it lacks the striking direction of an artist/great mind and averages out.
but most of what humans create isn’t novel, esp in saturated areas like art. so even LLMs can “just do that” they’re very capable at outputting worthwhile materisl
If you seriously dig into literature or film you'll find that many of the stories don't fit into the hero's journey at all--or only fit if you make the "journey" so generic as to be practically tautological.
As for the claim about people's lives. People who are serious about art almost invariably keep enjoying the "basic" classics; many musicians keep listening to Bach all their lives. Faulkner claimed to read Don Quixote over again every year. It is possible to find novelty in the well-worn.
Here's a scoop from inside writing circles (pre-AI era): we were a lot into "good writing techniques". Not into communication. We wanted to produce a "marketable product". On the way, we (writers, literature professors, critics, etc) trained readers and writers to appreciate technique by telling them there was such a thing. We established tropes, because "This enemies-to-lovers story is good" sells more than "you should read Book". AI is very good at techniques and tropes. If readers want authentic human writing, they will have to change.
It's quite likely that many survey participants were AI bots, and hence preferring AI generated content. And certaintly, that of the 1600+ people whose 15min they rented, few were paying any meaningful levels of attention to earn their 3 USD.
stories below. they were generated with chatgpt 4.0.
pair 1: https://osf.io/d8f6m/files/yjahv
pair 2: https://osf.io/d8f6m/files/sjph3
pair 3: https://osf.io/d8f6m/files/4pfw3
I cannot judge fairly given that the stories are labeled, but the story about the koi is dreadful. It is about a supposedly mysterious subject—the machine at least insists it's mysterious—but it is so dull, so pedestrian, so cut-and-dry. "My mommy dying makes me sad but the peaceful koi pond helps me find peace". Blech.
Perhaps it only seems so banal because the competition is so engigmatic.
And what does it mean for something to smell of "sterile detachment"?
Yes! Reading all samples really drives home the study's conclusion that “people generally prefer predictability, because difficult material requires more brain power.” I could blow through the AI generated versions as I do Brandon Sanderson or JK Rowling because my brain was saying "yada yada yada, yeah I get it, how many sentences do you need to say the same thing."
At first I was impressed by the “generated” stories and wondered about the prompt used to get such human sounding writing, except then I read further and saw they put the human written stories first and the AI stories after, which are all full of the familiar stylistic constructs (and banality) of AI written works.
They are not very representative. I absolutely hate the subjects and plots and writing all over and would have voted randomly.
Honestly much more curious to know about the human written stories were here! Feels at least pretty important to the study. With the AI, the specifics feel less important, because the relative input is going to be so small, so the researchers can probably just do it themselves. But how do you choose which works to stand in as pieces of human writing in general? How did they make sure the subjects weren't aware of the authors they chose?
"We assume creative writing requires uniquely human qualities, such as emotional understanding and lived experience, which leads people to underestimate AI's capabilities."
It's not about the relative "creativity" of it. Art is fundamentally about communication. It's a social activity, whether you're reading a book or listening to music or looking at a painting. Even a mid-average author is going to make something more compelling, on a human level, than extraordinary prose generated by a machine.
AI-generated creative artifacts are as interesting as the fractals in a snowflake. Interesting, sometimes beautiful, but void of human meaning.
Is art what is generated, or is art what is in the eye of the beholder? Do we find artistry in nature?
What does it say about us that the sum average of us disgusts us on a visceral artistic level, is that the machine's fault for reflecting us accurately?
Art is a social activity whereby a human uses their creativity to produce an artifact that expresses some aspect of their experience, using non-standard forms of communication. It's a way of signaling to others "do you see what I see? do you feel what I feel?" The beauty of art exists on two layers: the aesthetic quality of the work, and the sense of shared experience among those who consume the art.
We find aesthetic beauty in nature; we do not find artistry.
What was the quality? AI generally will have better grammar and such vs what most people write. So I could believe you prefer the AI content in this example because it is better quality, but this tells us nothing about AI vs human written when someone takes the time to edit their rough drafts.
(Note that hacker news discussions are not conductive to spending days crafting a good response - that takes days - so everything here is a rough draft)
Here are the stories, sadly, they immediately reveal the source so you can't play the quiz really. On the other hand, I think most people would spot the classics like em-dash and metaphores straight away. https://osf.io/d8f6m/files/osfstorage
The real question to be debated is how do you deal with a new kid on the block, who is better than you at [whatever].
Denial seems to be the general human default, it's easy and free, but at best it can only delay the inevitable.
Accept and adapt is probably the most logical approach, but that requires a lot of effort and risk taking, and still will likely leave you net worse off.
I suppose hostility and violence is another human favorite, but that almost never has gone over well.
This sounds as of AI has wider appeal. Like those bland mass-market beers that nobody likes very much but nobody hates either. Restaurants have those on the menu as a second choice.
While a real, published, author has taste and appeals to some people, not to everyone. Journal editors make sure of that.
These are ~1000-word stories (prompts at [1]), squarely in the sweet spot where planning logistics are minimal, context windows comfortably fit everything, and the appropriate level of abstraction omits any details which could glaringly reveal weak world-model knowledge.
SOTA LLMs from 2 years ago would saturate this study, and while modern LLMs are amazing and I suspect they would do quite well on a similar but more challenging study that forces them out of said comfort zone, this isn't that.
[1] https://www.cambridge.org/core/journals/judgment-and-decisio...
No, I have 0 interest in AI stories or music or art. I like all of these things because it's a view into the mind of the artist and I want to connect with another human and appreciate what they made and why.
I feel like this is just "we lied to people and we are surprised they trusted us" which just emphasize how hard we should work to enforce generated content as such (which I don't even know if it is possible in a reliable way)
Anyway, the good thing with human mind is it can explore, where a LLM is statistically producing words. It will certainly produce average stories, which if they are short will be better than most human writing. But it won't explore new ways of writing, which are the ones which can become a masterpiece IMHO
I would be interested in long stories experiments though, is the context wide enough to keep consistency and mid quality or are they becoming boring and un-readable like claude can be when trying to explain a concept ?
Then starts the philosophical questions : If a human need to write world rules, names, locations, characters, psychology, how each character perceive the world despite its rules, define relationship between characters and then delegate the story writing to a LLM, how much of the story belong to the human and what's the difference between doing this and hiring a ghostwriter
In short: creating stories requires mental effort, and humans don't want mental effort, so use AI.
It should be noted that these stories were 1000 word stories, with each one taking 5 minutes to read on average.
If it's not already the case, it likely will soon be that people prefer short main motifs for songs generated by AI over humans.
It could still be quite some time before they're writing entire songs and books, though.
No one buys 1000 word stories, or listens to 12 second songs on Spotify, so for the time being it's not really any cause for concern.
If Facebook has demonstrated anything it's that people have preferred AI memes for a while already.
People "prefer" lower quality art all the time - look at what's big on the radio and in theaters, what's popular on TV, etc.
Why should I be concerned if AI is making trash art instead of humans?
I loved Britney Spears when I was 9. Is the world really worse off if I'm listening to an AI version or Toxic or the Britney Original?
I can see why people would be scared if AI was pumping out Purple Rains and Superstitions on the regular, if it was writing a Secret History every hour, but it's not, nowhere even close.
This just seems like people being grumpy that people prefer Marvel movies over what they think are good movies.
Human artists have been pawning off derivative trash as art for a very very long time, and people have been consuming it and loving it. I'm not going to lose sleep if in 10 years AI is pumping out the derivative trash personalized to the exact level of trashiness desired by the consumer.
No one really knows how this will play out, but it's still uncomfortable to think about robots figuring out how to reliably make things that look good, sound good and are full of meaning, would people still have any motivation to create? Would anything be considered good if everything is "good"? Of course things could turn out fine, I just wouldn't minimize the significance of what's happening.
This is already the struggle for 99% of people that would like to be artists (or athletes, or a number of things)...
>It should be noted that these stories were 1000 word stories
A huge limitation of the study IMO.
Most people who read are not reading 1000 word short stories. If they could replicate this story using novella length stories (25-40k words), I'd be curious to see the results. I would wager that a much larger % of people would prefer long form content from humans.
TDR: no.
Funny. It's the same result as that famous painting labeled as AI-generated and people found many faults with it. It's just showing that people are biased against AI itself; the quality of the content doesn't really matter.
LLM writing is optimized, at a great expense, for that specific parameter. They literally have human raters fine-tuning "favorability" - general appeal - as a part of post-training. So yes, the conclusion here doesn't surprise me at all. By the way, this is also true of HN! You think you're better than that, but at this point, on any given day, we have around 10 LLM-generated stories on the front page (and another 10 are usually AI-related press releases or op-eds).
Anyway, human writing isn't like that. You have a style that some people will like, some people will find "meh", and some people will hate.
A better question is, do people want to read AI text all day? And the answer to that question is, I think, very different. Most people don't. Which is why so many parties are really determined not to tell you.
No. Dear lord no.
One thing I am getting really tired of is the growing inability to distinguish legitimate sources. I don’t want to have to play, “can you tell the difference?” It’s not fun being had by a liar.
I want human written content or nothing at all. I care about the struggle. I like the spelling and grammar mistakes. I enjoy the story between the lines. I like the growth of the author and watching them and their work evolve.
Generated fiction is boring because it has none of that. There is no life in it. No author.
“A human wrote it,” is a light trick. Ok, who? Where are they from? If you have to lie to give answers in order to convince someone they like the story you’re not proving that people prefer AI stories in my estimation: you’re proving that the test is a convincing lie. And maybe that people are gullible.
I'll save you the time and effort. No.
Stories are always based on realities experienced by the authors, either personally, or viscerally. AI has no experience of this kind.
The experiences then are assembled or re-assembled into an art form, that makes impact on the thinking reader. The key ingredient is how this experience was meaningful, useful, or otherwise impactful enough to have qualified to become art.
Thus, reading fiction (good ones, not the rubbish kind) lets us experience those moments without the actual life-level cost (the only cost is reading). And therefore, these experiences get transferred to us, making us wiser, more aware, more rational-- if the art is real art.
Take the experience-as-art out of the equation and all is left is just slop.
Very bleak. I think like most people, I would prefer for stories not to be written by AI. I value human creativity, and believe that an LLM cannot truly produce art.
This article suggests that when people are lied to that stories were not written by AI (when they actually were), they prefer them.
For practical questions, I always prefer Claude's answer to a human's. Non-narcissistic and to the point. For entertainment, I'm less impatient, although unsurprisingly, most people view reading as homework and are happier to blow through a book written at a YA level. Brandon Sanderson a'int the king for nothing.
"Art is in the eye of the beholder" - it is me, who enjoys art and how it affects me and works with my emotions that is affected by it. Who or what the artist is, should be irrelevant, unless the artist imposes some monetary value I want to profit from at some point.
I don’t think the artist should be irrelevant - the art is the expression of the artist’s emotions. Knowing about the artist unlocks a lot of depth in the art, it’s an attempt to share and express their perspective through creative means.
Not to say how you decide to consume art isn’t valid (there is no proper definition after all). However, you’re doing a disservice to a lot of artists by dismissing their role in it all.
> However, you’re doing a disservice to a lot of artists by dismissing their role in it all.
That is a false assumption. I do not dismiss the artist role, but I also would not be interested in the artist if their art does not move me. People generally do not know who Banksy is, but he is being defined by his art, not by his persona. That may very well happen for an AI agent sometime in the future.
It does - but does this not potentially suggest that an LLM can produce "art"?
To me "art" is the communication of something from the artist, through arrangement of reality, to the person experiencing the art.
Define art?
Yeah why not?
Srsly.
Nonethless, a lot of people enjoy stupid movies, others enjoy smart movies etc.
If you are not like me and you don't care about the time you spend on a 2h movie because you just want to watch something, you will have a complete different experience than me. I watched a 1000 movies and more, I want/I need something 'new' or unique i haven't seen a 1000 times.
Its the same with the AI slop videos; I found the first 3 'cats with jobs ai' videos funny as a category, i don't want to watch thou 'accidents' videos if they are fake.
I'm sorry but all this proves is that a lot of published stories aren't very good. The three human stories attached don't have fantastic plot and read a little poorly.
All three of the human written stories read like they were made for a reviewer/publisher rather than a reader.
I don't think this is a good study or a good representation of good writing. I think many short stories have the issue of force feeding so much information and tricks that the 3 human stories read more gimmicky than the chatgpt ones.
It's pseudoscience no matter what, but I'd like to see the results sliced up for an age range of roughly 35-70. I would heavily question the taste and judgment of those outside that range. Or at the very least note that people's personalities will change significantly when entering/exiting that range.
That's why those "we prefer human writing" polls (such as the one cited in [0]: whooping 98% readers say they prefer human writing to AI!) are just noise.
To quote myself:
> People are really, really good at lying to themselves, let alone to an online poll. They'll tell you that they prefer imperfect or even bad writing as long as it's not AI slop, just like how they'll tell you they like healthier food, they prioritize personality instead of look for potential dates, how they use LLM "only as a spellchecker", and how they use tiktok for educational videos. As long as there is no stake, people will just say what make they feel better.
[0]: https://bcantrill.dtrace.org/2026/09/05/the-revolt-of-the-re...
> That's why those "we prefer human writing" polls (such as the one cited in [0]: whooping 98% readers say they prefer human writing to AI!) are just noise.
Why would they not?
> They'll tell you that they prefer imperfect or even bad writing as long as it's not AI slop,
But I do, and it's very clear that the online communities I frequent prefer people write in their own voice, even with very bad grammar and spelling, rather than resort to chatgpt for their posts and comments — this comes up all the time
> just like how they'll tell you they like healthier food
But lots of us do? Once you are over 30 and you control your own fridge and cook your own food, why would you not?
> they prioritize personality instead of look for potential dates
But (in a fair context for comparison) people actually do, especially women. As long as someone is presentable and shows a bit of self-care and self-respect, for the most part looks rank at least on a par with personality, if not below it.
> how they use LLM "only as a spellchecker"
I don't use it at all for writing; I do know people are weak on this point but I also know that people understand it fosters distrust
> how they use tiktok for educational videos
Huge chunks of people don't use it at all.
Most of what you are saying here sounds like you are projecting low mood and your own personal prejudices and insecurities. Try to be less misanthropic.
Of course more than zero people actually use tiktok for educational purposes and like healthy foods. I'm talking about how questions like that ("do you prefer imperfect human writing or AI writing?", "what do you use tiktok for?", "do you stop reading if you know something is written by LLM?") are ill-designed for questionnaires.
Tests like the OP's link are slightly more meaningful to find out people's true preferences. But of course the best way is still A/B test and tracking user behavior. Anyone has done any form of behavioral tests, online or not, would understand that what people actually do is vastly different from what they think they do. The correlation is often negative.
"Do you prefer imperfect human writing or AI writing?" is a question that is loaded.
It's absolute horseshit (and I would say somewhat sociopathic) to try to A/B test that to catch people out as if to say "aha you are lying to yourself because you can't tell".
People's preferences are representative of their desires and aspirations, their future self-perception, their better self, etc. I'm sad that you seem to think that is something they should be disabused of by statistical means.
It would be misanthropic to assume that because people say they want healthier food but then don't buy it, that they don't really want it. The proper response to that is not to give people what their behaviour says they will actually buy; it is to broadly compel food producers to make healthier food that matches what they wish they would more often choose.
ETA: above all what I want to impress upon you is that "people are really good at lying to themselves" is not actually a statement of fact; it is your opinion and interpretation projected onto the data you choose. There are other ways to explain this gap that do not involve fundamental attribution error.
This is a bit like asking "do people prefer vegetarian meals if we don't tell them there's actually meat in it?"
People care about the production process even if they can't tell you're lying to them about it.
Of course AI is hyped at Villanova:
https://www.villanova.edu/university/ai.html
"Villanova embraces Artificial Intelligence (AI) with a commitment to thoughtful, ethical use rooted in our Augustinian Catholic mission and values. We aim to balance innovation with our humanist tradition and intellectual heritage, preparing students to lead responsibly in a changing world.
Explore resources designed to help you learn, teach and work with AI at Villanova."
What does that have to do with the results I'd the study? Do you think there will be a different result if it was done by anyone else, anywhere else?
Yes, with a proper story selection you can get any desired result if you repeat the study.
Sure studies can be designed to get a desired result, but we're talking about this particular study. Do you see any issues in how it was done that would point to designed-for-desired-result?
Rare exception of Betteridge's law of headlines
No
Reading this, I'm reminded of... New Coke.
Despite post facto conspiracy theories that it was a guerilla marketing campaign to promote "Old Coke", it was not. Coca-Cola tested a new formula in focus groups and people preferred it at the time because, basically, it was sweeter. But this didn't translate into longer term consumption. It's a cautionary tale in how data science can fail and how you shouldn't necessarily extrapolate to the long-term from the short-term.
So when people say they preferred AI written stories, what are they really reacting to? There's probably a lot of elements that are engaging but is it a lasting story? Most enduring stories become so because they speak to some truth of the current times or the human condition that can be hard to separate from the technical merits of the story.
Example #1: Fight Club both as a book and a movie was a bit of a sleeper hit. It was a box office failure. But it became a cult classic because it was one of the earlier works that spoke to the loss of a future for men in our society.
Example #2: The Rocky Horror Picture Show. Again, it was a sleeper hit but it built over time as a relatively early Western expression of queer culture, gender identity and cultural conformance.
Example #3: The Wire. While it had great writing it has endured as a testament to institutional failure at every level and that doesn't necessarily take a villain. Everybody can be acting in their own interests and you end up with inevitable failure anyway.
Example #4: The Great Gatsby. This was never a hit in F. Scott Fizgerald's lifetime. How it became one is an interesting story. Basically, during WW2 the Army created these small print books for deployed soldiers to easily carry and read. Then once the war ended and many of the veterans went to college, they wrote about the Great Gatasby. It's now viewed as an American masterpiece that captures the 1920s like few other works. It expresses class consciousness and some of the metaphors (like the green light on the dock) are debated to this day.
Now, could AI write things that coherently speak to the human condition this way? Not yet (IMHO). Far from it.
That being said, there are only so many base story structures. Take the romance category. You'll find repeated tropes. Could AI write what might best be called "pulp fiction" in that genre? Probably sooner rather than later.
No