I've asked ChatGPT to write some stories, and I have found the output amusing. I could see reading some of them to children. Maybe they don't make a great deal of sense, but they're still fun. Or perhaps, because they don't make a great deal of sense, that's what makes them fun?
As great as ChatGPT is, the content it generates tends to follow some fairly repetitive patterns & styles, and it's going to be pretty easy to identify unless heavily edited.
However, I also suspect this technology will develop very rapidly: just look at how quickly the AI art scene has evolved since Stable Diffusion came along. Pretty crude at first, but the quality of stuff coming out of it is amazing now.
I feel like the biggest impact will be for authors who have an existing body of work... you can train the model on your writing and generate new text in your own style. What takes days could be done in hours.
I'm sure the same could be done for editors... train it on annotated text (original + corrections), use it to generate new edits.
Sure, but most authors aren't payed for their writing style, they are paid for what they say, for a unique insight, good research, or, if fictional, for a good story arc and suspense. I haven't seen purely AI written pieces that have that. And I am not sure they will with the current approach, which produces mostly bland and generic text. We would need an AI that does some actual creative thinking first.
I think (hope) that AI, in the interim, will provide actual authors ways to get inspired by its random nature. Like painters might start with an ink blob and then paint over it. Or let AI fill in some gaps, like description of people or places, etc.
(Copying my comment over from a different HN ChatGPT thread I commented on this morning - are all these ChatGPT threads getting a bit ... samey?)
As a writer, I'm much happier in the revision/editing stage than in the first draft stage. I have a half-finished novel that's been half-finished for the best part of 8 years now. I'm now thinking that maybe if I feed the draft into <ML-shiny-of-the-day> with suggestions for how I want the next section/chapter to develop, it could generate some draft copy for me. Possibly several different versions. After which I'm in my Happy Editing Place, shaping existing copy to something I like. Repeat and rinse!
The end result would be 100% my work as the ML algorithm/brains/whatever would be generating new copy based on my existing copy and my vision and direction for the novel. And I get the final say on the results. Win-Win!
There are services to identify LLM output, like GTPZero and a classifier from OpenAI, but they aren't very reliable even at this early stage in the development of LLMs—lots of false positives and negatives. They'll likely be useless when the tell-tale signs they rely on are ironed out.
If you separate generating the content from stylizing it you can make it much more natural and targeted toward your readers. This generation of models can do a lot but does things best when you do one thing at a time.
I don't think that's the right term. Eternal September refers to too big of an influx of new people for the current members to inculcate with the local manners, morals, and attitudes resulting in a degradation of an existing culture. Amazon authors are already too diverse and dispersed to have a singular culture. This is more like mass production meets content creation or an assembly line for knowledge work.
>Ghostwriting - by humans - has a long tradition, she said, but the ability to automate through AI could turn book writing from a craft into a commodity.
I think writing has already become a commodity. Go on Fiverr and you'll find people offering to write on any topic, fiction and non, for fractions of a cent/word.
The biggest reason I can see someone using ChatGPT to generate a children's book vs paying someone overseas to do it is about $10. Words are cheap, astonishingly cheap, what's not cheap is reputation and credibility. I don't see a big reason for authors to worry, yet.
I agree, I did much ghostwriting to make money in grad school. A lot of the books I worked on for other people were them exploring a new revenue stream and capitalizing on their existing reputation. The words themselves are not really what sells the work.
People looking to get rich quick write (or pay sweatshops to write) very low effort, poorly researched books about obscure topics. They put them on amazon, maybe make some money, then repeat. The video explores one of those buyable programs to get you involved in all of that.
Seems like chatGPT is being used as the next tool in this process
The killer feature the publishing industry and writers bring to the market is curation. That is curation of the facts, story line for brevity, etc. all things ChatGPT cannot do.
This also is a problem though because tools like it can be used to DOS this curation - see e.g. todays Clarkesworld issue (https://news.ycombinator.com/item?id=34875781) - this can be worked around by only curating from known sources... which I can imagine provides a massive barrier to new people.
And of course, if anyone makes something new it'll just be instantly ripped off and redistributed via copyright laundering.
It's hard not to view this as enforced enshittification for everyone, all for the benefit of a few quick bucks for massive corporations.
Shouldn't there be a "massive barrier" to new people writing checks notes AI-generated children's books?
I've read now to 4 kids and the vast majority of children's books are terrible. A few are evocative, imaginative, have memorable characters, interesting, thoughtful messages - but most of them are a waste of paper and time.
Most children's books - even the ones more deliberately planned than a ChatGPT textdump - are very lazily conceived and executed.
You should probably check those notes again. I'm saying that one of the easiest and probably most effective responses to this by publishers is to only use known sources e.g. existing authors and people they can vouch for.
This would put up a massive barrier for AI-generated books, yes - as intended. But also puts up a massive, massive barrier for finding new _real_ authors, who don't have the connections and might have previously actually had a chance of being published.
With very few exceptions, I have not seen publishers doing that over the past few decades. I see no brevity & the curation of facts is a topic for another day....
So, he used ChatGPT to create a children's book in _4 hours_ and he is now a published author, the very thing he dreamed to become? Please...
"The idea of writing a book finally seemed possible"
There is no mention of the guy being disabled or unable to write himself. It's a complete fluff piece that Reuters failed to properly articulate. They should've started with the Amazon part and then used this guy as an example rather use a pretentious introduction that deserves no praise nor any respect.
> So, he used ChatGPT to create a children's book in _4 hours_ and he is now a published author, the very thing he dreamed to become? Please...
Well he is going to find out that his 'books' in the US are essentially in the public domain and are un-copyrightable due to this ruling: [0]
Human authorship is required for obtaining copyright for the books in the US and not when an AI has done the work entirely or has done the majority of the work in both generated text and images and especially when they are crediting an AI as the author or co-author.
I wonder whether this will change when we get self-aware sentient AIs. What if e.g. Koko the gorilla [0] had made a painting? Could they have claimed copyright with their limited sign language capability?
Content got an order of magnitude cheaper to create so now nobody cares about their copyrights and at the same time nobody cares about ripping it off. Strange times.
Possibly. Possibly not. In the case where a money took a selfie, it was ruled as uncopyrightable[1]. Just because the photographer had set the stage, the work completed was done my by something unable to claim copyright. The better solution is to not acknowledge that ChatGPT was the source. After all, it won't complain.
I've tried a couple of apps that promise to create children stories based on ChatGPT and while it works (-ish) for short stories I think the only way to get a better and longer output that is decent enough to warrant publishing is to be very hands on with the prompts, at which point I'd argue it's possibly considered real creative work and ChatGPT is just a (very powerful) tool, not a full-blown substitute for a writer/author.
I’m not so worried about books, because there’s generally a human in the loop between writing a book and publishing it, but how long do we expect search engines to remain useable? Blogspam is already a real issue when it comes to finding stuff via search, I have a bad feeling it’s going to get much harder to find what you need to find as machine learning tools out-SEO humans.
It's very easy to get low content book published through Amazon's Direct Publishing platform. I don't really think it's a problem personality, but if you're opinion is premised on publishing being difficult I think it's misplaced.
> There were over 200 e-books in Amazon’s Kindle store as of mid-February listing ChatGPT as an author or co-author, including "How to Write and Create Content Using ChatGPT," "The Power of Homework" and poetry collection "Echoes of the Universe." And the number is rising daily. There is even a new sub-genre on Amazon: Books about using ChatGPT, written entirely by ChatGPT.
Given that these books are published in the US, good luck attempting to copyright something that has been majority outputted or entirely created from an AI like ChatGPT or Stable Diffusion since that is completely un-copyrightable and is instead public domain. [0]
Some nerd on here called it a human "stranglehold" on creativity and art and it's one of the funniest most brain-dead things I have read about AI. It's like so many of these programmers out here can only see programs and tech, and have completely lost touch with their humanity. Another even went so far as to say that, given enough billions of dollars, AI will be churning out art as good as anything fine arts. It's an unbelievably funny take
Humanity. Hm. Maybe. But is it possible that some who seek to limit what "humanity" means are also gatekeeping a safe place for themselves while other humans want to move forward and explore? Humans, uniquely, are explorers and triers of crazy new ideas. Is that not an essential part of humanity and art?
IMO, your "nerd" and "brain-dead" insults are a bit immature.
I'm not praising any current AI-generated art. It's so early. Let's see where this goes. If it all falls flat in a few years then the experiment was still essentially a human thing that had to be done IMO.
If you (royal you) want to pass off AI prompting as equal to real human expressions of pain, love, anger, et al. you are free to do so. Nobody is stopping you!
Edit: Actually, I would even go so far as to say that the prompt itself, that you a person wrote, to generate whatever, is the real artistic expression. Don't even AI require output!
The kind of word you’d expect to come out of the mouth of an A16Z VC shilling their latest hot AI startup.
“Today nearly all creative output comes from creative humans and those humans capture an appallingly high 25% of the economic value. With CreAItive the human stranglehold on creativity has been escaped, 100% of CreAItive’s output is owned by investors.”
Looks like click farms have expanded into book publishing. The tech revolution of diminishing standards seems to be progressing steadily. You can now ride in your Cruise self-driving car that drives slow and stops in awkward places, while reading a soulless, irrelevant book and look up inaccurate info in Bing. The future really is never as you picture it.
This has been going on for a while. They refer to it (themselves) as low-content publishing and it's popular among the same crowd as the content marketing/low-content website ad farms. Before ChatGPT people used tools like Writter AI and Jarvis AI (now Jasper AI), but the newer models are much better and incorporate their own, possibly incorrect, knowledge.
So, Grandma got my little monster a "custom book" recently. It's got monster's name, and friend's name, and grandma's name of course.
It's so awful, that it's really hard to believe that it's just a poor translation. It really only makes sense if it was written by an AI. But I question why... It's just not that hard to write a 14 page story about unicorns and special talents.
It seems like more work to have an AI spit it out. It's just variable names for CHILD, FRIEND1, FRIEND2, etc. So, I think it was AI written, but nothing about it makes sense.
The next wave of print journalism is upon us. Soon we will gladly companies to get access to their organically created and edited content slathered in Ads and DRM.
I just went down a rabbit hole for the last 30-60 minutes , looking at some of your videos (like your camera setup for your desktop) and your linkedin (impressive), lol. you got some interesting stuff up
Well thank you kindly sir. I had a lot more videos on game development, mathematics, programming, electronics tinkering and a bunch of other stuff at one point in time. Unfortunately YouTube is a cesspool of individuals taking other people's videos, slapping their logo on it and reuploading. I decided I didn't want to play that game anymore.
I want to hate on ChatGPT-written books... But then there's this.
My Prompt (I have endless random ideas like this as I walk the dogs and listen to audio books):
Suggest the plot for a military science fiction book. The galaxy is divided into factions based on their fundamental properties: organic vs. artificial, living vs. dead, good vs. evil, plant vs. animal vs. mineral.
ChatGPT:
Title: "Galactic Rift: The War Between Factions"
In a faraway galaxy, a vast and complex network of planets is divided into factions based on their fundamental properties. The organic and living species have formed an alliance, while the artificial and dead ones have joined forces to control the galaxy. The plant, animal, and mineral factions are split amongst the two alliances, leading to a tense political climate.
Amidst this divide, a brilliant military strategist, General Alara, commands a flagship of the organic alliance, the Stellar Phoenix. She is tasked with leading her crew through dangerous space battles against the ruthless and technologically advanced artificial faction. The stakes are high, as the victor will control the galaxy and its resources.
As General Alara and her team engage in war, they encounter unexpected allies and enemies. They discover that not everything is as it seems, and that the factions are more complex than their fundamental properties suggest. The lines between good and evil are blurred, and each side is driven by their own motivations.
As the war intensifies, General Alara must make difficult decisions that challenge her beliefs and loyalties. She is forced to confront the dark side of her alliance and question whether the organic faction truly represents the good side of the war. Her personal journey is intertwined with the fate of the galaxy, as the war reaches its climactic conclusion.
In "Galactic Rift: The War Between Factions," the reader is taken on a journey through a galaxy filled with political intrigue, space battles, and philosophical questions. The plot explores themes of identity, morality, and the consequences of war. The novel offers a unique take on military science fiction, breaking free from the traditional tropes and challenging the reader's assumptions about good and evil.
It came back with 550 words, all perfectly OK and pretty much cogent, but it would suck as a chapter of an actual book.
Generally, it had at least two fatal flaws:
(1) it was all cliche. I don't really mind cliche stuff for my walking-the-dogs audiobooks, but it can't be all cliche. There needs to be something in there.
(2) it didn't delve into any details. E.g. there was an important planet, unexpectedly under attack, and a space battle for it. But nothing about why it was important and the space battle was over in a sentence or so. I prompted for more detail on these and got some, but again, it was purely cliche and contained no interesting details.
Also, there was literally nothing following from the dynamic the premise suggests.
With a bunch of prompting I think I could get it to write the book, but I think it would be purely cliche and bland, except possibly where I did some heavy lifting.
Ha, well, considering how it works, I would be interested if it wasn't.
I think things like "Write me a song about a dumper baby's birthday in the styles of Shakespeare and Tupac" work well not because they are terribly deep into any aspect, but two cliche takes wrapped up into something mediumly unique instantly.
I've had fun doing bedtime stories by having ChatGPT create a book about a subject & a character names we agree upon. It's a great way to get into reading. In a way, it's similar to choose your own adventure style books. We haven't brought SD into it yet but that would be a fun addition to create images that go with the book.
I've also used ChatGPT to poke fun of family members in a Dr Seuss style short story. Great laugh for the whole family as the short stories have been really good. And people love hearing their names in stories.
I think this gives more people tools to easily create books. Hopefully it improves literature. Much of the books I've read that have been produced in the past few decades have a lot of fluff. Same with video series on Netflix, etc... Plenty of stories created that don't add to character development or the story. Plenty of pages written that don't add to a non-fiction book's value but are meant to add pages to meet a publisher's criteria & come off worse than a college kid increasing font & line spacing.
It'll be interesting to see how these stories compete & which stories get the most recommendations & "authentic likes".
I spent a solid hour with chat GPT writing a story about a guy who meets a talking cat that leads him to a magical word within our world (sort of like the wizarding world being hidden in plain sight in Harry Potter) and it was tons of fun.
It felt like the best "story where you are the hero" ever.
A book is more than 5 short paragraphs of synopsis, especially good books that are worth re-reading. Ideas are cheap and sketching a framework is the easiest part in building something. Fleshing out the details is a task that currently available LLMs aren't yet capable of doing.
Dune can be summarized in similar lengths too, but that's not what made the series a popular sci-fi universe.
ED Hirsch is looking more and more prescient every day.
I mean, really, sure: there are a few dozen really clever children's books that come out every year that deserve to be added to the "pantheon" but isn't this area basically saturated?
Imagine you read your child one book a day for their first 5 years, you and they co-read two books a day from age 5 to age 10 and then they move on forever to whatever arbitrary boundary we set for "children's books" versus the rest of literature.
Assuming a modest amount of re-reading your favorites, that's an ambitious local maximum of about 5,000 books.
Realistically, this number is more like 1,000 to 1,500 books. This space is already well curated - Scheherazade has you covered in a single compendium!
There's no way any random ChatGPT production is going to cut into the space.
85 comments
[ 4.7 ms ] story [ 167 ms ] threadHowever, I also suspect this technology will develop very rapidly: just look at how quickly the AI art scene has evolved since Stable Diffusion came along. Pretty crude at first, but the quality of stuff coming out of it is amazing now.
I'm sure the same could be done for editors... train it on annotated text (original + corrections), use it to generate new edits.
I think (hope) that AI, in the interim, will provide actual authors ways to get inspired by its random nature. Like painters might start with an ink blob and then paint over it. Or let AI fill in some gaps, like description of people or places, etc.
As a writer, I'm much happier in the revision/editing stage than in the first draft stage. I have a half-finished novel that's been half-finished for the best part of 8 years now. I'm now thinking that maybe if I feed the draft into <ML-shiny-of-the-day> with suggestions for how I want the next section/chapter to develop, it could generate some draft copy for me. Possibly several different versions. After which I'm in my Happy Editing Place, shaping existing copy to something I like. Repeat and rinse!
The end result would be 100% my work as the ML algorithm/brains/whatever would be generating new copy based on my existing copy and my vision and direction for the novel. And I get the final say on the results. Win-Win!
https://gptzero.me/
https://openai.com/blog/new-ai-classifier-for-indicating-ai-...
I think writing has already become a commodity. Go on Fiverr and you'll find people offering to write on any topic, fiction and non, for fractions of a cent/word.
The biggest reason I can see someone using ChatGPT to generate a children's book vs paying someone overseas to do it is about $10. Words are cheap, astonishingly cheap, what's not cheap is reputation and credibility. I don't see a big reason for authors to worry, yet.
https://youtu.be/biYciU1uiUw
People looking to get rich quick write (or pay sweatshops to write) very low effort, poorly researched books about obscure topics. They put them on amazon, maybe make some money, then repeat. The video explores one of those buyable programs to get you involved in all of that.
Seems like chatGPT is being used as the next tool in this process
And of course, if anyone makes something new it'll just be instantly ripped off and redistributed via copyright laundering.
It's hard not to view this as enforced enshittification for everyone, all for the benefit of a few quick bucks for massive corporations.
I've read now to 4 kids and the vast majority of children's books are terrible. A few are evocative, imaginative, have memorable characters, interesting, thoughtful messages - but most of them are a waste of paper and time.
Most children's books - even the ones more deliberately planned than a ChatGPT textdump - are very lazily conceived and executed.
This would put up a massive barrier for AI-generated books, yes - as intended. But also puts up a massive, massive barrier for finding new _real_ authors, who don't have the connections and might have previously actually had a chance of being published.
"The idea of writing a book finally seemed possible"
There is no mention of the guy being disabled or unable to write himself. It's a complete fluff piece that Reuters failed to properly articulate. They should've started with the Amazon part and then used this guy as an example rather use a pretentious introduction that deserves no praise nor any respect.
Well he is going to find out that his 'books' in the US are essentially in the public domain and are un-copyrightable due to this ruling: [0]
Human authorship is required for obtaining copyright for the books in the US and not when an AI has done the work entirely or has done the majority of the work in both generated text and images and especially when they are crediting an AI as the author or co-author.
[0] https://www.copyright.gov/rulings-filings/review-board/docs/...
[0]: https://en.wikipedia.org/wiki/Koko_(gorilla)
There will be no need for Microsoft, for OpenAI, for copyright, for software, for open source, for jobs. It will be a civilisation shattering event.
I think it’s already going to be fun watching OpenAISoft stop their own products being used to recreate themselves.
“Write my a functionally identical piece of software to Microsoft Word”.
People will get around it saying no and open source alternatives will eventually be available anyway.
It’s really incredible what’s going to happen.
[1] https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...
Given that these books are published in the US, good luck attempting to copyright something that has been majority outputted or entirely created from an AI like ChatGPT or Stable Diffusion since that is completely un-copyrightable and is instead public domain. [0]
[0] https://www.copyright.gov/rulings-filings/review-board/docs/...
Humanity. Hm. Maybe. But is it possible that some who seek to limit what "humanity" means are also gatekeeping a safe place for themselves while other humans want to move forward and explore? Humans, uniquely, are explorers and triers of crazy new ideas. Is that not an essential part of humanity and art?
IMO, your "nerd" and "brain-dead" insults are a bit immature.
I'm not praising any current AI-generated art. It's so early. Let's see where this goes. If it all falls flat in a few years then the experiment was still essentially a human thing that had to be done IMO.
Edit: Actually, I would even go so far as to say that the prompt itself, that you a person wrote, to generate whatever, is the real artistic expression. Don't even AI require output!
“Today nearly all creative output comes from creative humans and those humans capture an appallingly high 25% of the economic value. With CreAItive the human stranglehold on creativity has been escaped, 100% of CreAItive’s output is owned by investors.”
It's so awful, that it's really hard to believe that it's just a poor translation. It really only makes sense if it was written by an AI. But I question why... It's just not that hard to write a 14 page story about unicorns and special talents.
It seems like more work to have an AI spit it out. It's just variable names for CHILD, FRIEND1, FRIEND2, etc. So, I think it was AI written, but nothing about it makes sense.
"Written by Artificial Intelligence" 2,000 results.
https://justinlloyd.li/blog/the-greatest-story-never-sold/
My Prompt (I have endless random ideas like this as I walk the dogs and listen to audio books):
Suggest the plot for a military science fiction book. The galaxy is divided into factions based on their fundamental properties: organic vs. artificial, living vs. dead, good vs. evil, plant vs. animal vs. mineral.
ChatGPT:
Title: "Galactic Rift: The War Between Factions"
In a faraway galaxy, a vast and complex network of planets is divided into factions based on their fundamental properties. The organic and living species have formed an alliance, while the artificial and dead ones have joined forces to control the galaxy. The plant, animal, and mineral factions are split amongst the two alliances, leading to a tense political climate.
Amidst this divide, a brilliant military strategist, General Alara, commands a flagship of the organic alliance, the Stellar Phoenix. She is tasked with leading her crew through dangerous space battles against the ruthless and technologically advanced artificial faction. The stakes are high, as the victor will control the galaxy and its resources.
As General Alara and her team engage in war, they encounter unexpected allies and enemies. They discover that not everything is as it seems, and that the factions are more complex than their fundamental properties suggest. The lines between good and evil are blurred, and each side is driven by their own motivations.
As the war intensifies, General Alara must make difficult decisions that challenge her beliefs and loyalties. She is forced to confront the dark side of her alliance and question whether the organic faction truly represents the good side of the war. Her personal journey is intertwined with the fate of the galaxy, as the war reaches its climactic conclusion.
In "Galactic Rift: The War Between Factions," the reader is taken on a journey through a galaxy filled with political intrigue, space battles, and philosophical questions. The plot explores themes of identity, morality, and the consequences of war. The novel offers a unique take on military science fiction, breaking free from the traditional tropes and challenging the reader's assumptions about good and evil.
Not bad. I'd read that book.
Let's see how well the AI actually explores those themes.
It came back with 550 words, all perfectly OK and pretty much cogent, but it would suck as a chapter of an actual book.
Generally, it had at least two fatal flaws:
(1) it was all cliche. I don't really mind cliche stuff for my walking-the-dogs audiobooks, but it can't be all cliche. There needs to be something in there.
(2) it didn't delve into any details. E.g. there was an important planet, unexpectedly under attack, and a space battle for it. But nothing about why it was important and the space battle was over in a sentence or so. I prompted for more detail on these and got some, but again, it was purely cliche and contained no interesting details.
Also, there was literally nothing following from the dynamic the premise suggests.
With a bunch of prompting I think I could get it to write the book, but I think it would be purely cliche and bland, except possibly where I did some heavy lifting.
Ha, well, considering how it works, I would be interested if it wasn't.
I think things like "Write me a song about a dumper baby's birthday in the styles of Shakespeare and Tupac" work well not because they are terribly deep into any aspect, but two cliche takes wrapped up into something mediumly unique instantly.
I've also used ChatGPT to poke fun of family members in a Dr Seuss style short story. Great laugh for the whole family as the short stories have been really good. And people love hearing their names in stories.
I think this gives more people tools to easily create books. Hopefully it improves literature. Much of the books I've read that have been produced in the past few decades have a lot of fluff. Same with video series on Netflix, etc... Plenty of stories created that don't add to character development or the story. Plenty of pages written that don't add to a non-fiction book's value but are meant to add pages to meet a publisher's criteria & come off worse than a college kid increasing font & line spacing.
It'll be interesting to see how these stories compete & which stories get the most recommendations & "authentic likes".
It felt like the best "story where you are the hero" ever.
Dune can be summarized in similar lengths too, but that's not what made the series a popular sci-fi universe.
I mean, really, sure: there are a few dozen really clever children's books that come out every year that deserve to be added to the "pantheon" but isn't this area basically saturated?
Imagine you read your child one book a day for their first 5 years, you and they co-read two books a day from age 5 to age 10 and then they move on forever to whatever arbitrary boundary we set for "children's books" versus the rest of literature.
Assuming a modest amount of re-reading your favorites, that's an ambitious local maximum of about 5,000 books.
Realistically, this number is more like 1,000 to 1,500 books. This space is already well curated - Scheherazade has you covered in a single compendium!
There's no way any random ChatGPT production is going to cut into the space.