AI can create pictures and music from prompts, how soon will it be before they can create novels?

[CultureCitizen]

You can already create plots using chat-gpt.
It can even write cheese first-person stories if prompted. Or pieces of stories.

How long do you think it will pass before they can create a full novel by themselves?
What place will writers have when this happens? Will writing be only a hobby?

On the bright side some sf scripts for tv and cinema are dreadful and might actually improve.

 

[BAYLOR]

You can already create plots using chat-gpt.
It can even write cheese first-person stories if prompted. Or pieces of stories.

How long do you think it will pass before they can create a full novel by themselves?
What place will writers have when this happens? Will writing be only a hobby?

On the bright side some sf scripts for tv and cinema are dreadful and might actually improve.

I can't disagree with that last bit, AI's probably would do better with tv and movie scripts.

 

[Steve Harrison]

Probably not long. As soon as publishers can request a genre x novel in the the style of bestselling author y, writers will be gone.

I'd like to think that's not the case, but when I consider books like the Jack Reacher novels (which I thoroughly enjoy), I reckon AI could already produce one that would be hard to distinguish from the real thing.

Hope I'm wrong!

 

[ckatt]

As a reader, I wouldn't be interested in books by an AI except for maybe the "novelty"
My bigger concern would be about people trying to pass off AI writing as their own since AI generated works can't be copyrighted.
Film might be different though. As much as a good script matters it can be make or break based on the actors performance. And that's what most people are interested in when it comes to movies. The actors, not the script writing team.

 

[CultureCitizen]

Probably not long. As soon as publishers can request a genre x novel in the the style of bestselling author y, writers will be gone.

I'd like to think that's not the case, but when I consider books like the Jack Reacher novels (which I thoroughly enjoy), I reckon AI could already produce one that would be hard to distinguish from the real thing.

Hope I'm wrong!

There are novels and even short stories that have layers. You can enjoy the action or the drama of the novel on the top layer. Then you can think about the actions and the moral choices and character arcs on a deeper level. On an even deeper level you can even wonder what gave birth to thee situation in a novel ( when I read 1984 I was constantly wondering what would be the motivations to create such a distopic society... a question that remains unanswered, because Hitler , Stalin , Pol Pot and Kim Jong Un have managed to create very similar societies).
And yes, I am wondering if AIs will ever achieve such a deepness of thought.

 

[Harpo]

Alexa, re-write “Therese Raquin” and “The Murder Of Roger Ackroyd” into a single novel set in Ankh Morpork but with scenes from The Goon Show (especially “Six Charlie’s In Search Of An Author”) and spaghetti westerns. Include a subplot about twin macguffins and cake.

Easy peasy.

 

[cyprus7]

Perhaps it will write the book and the movie (starring you, of course.)

A quick and sobering guide to cloning yourself

It took me a few minutes to create a fake me giving a fake lecture.

oneusefulthing.substack.com

 

[sknox]

When will it happen? Now. Basic rule of thumb: if you are thinking about the future, it's probably already passed you by.

What will be the place of writers? Heck, writing for most people isn't profitable anyway. It's not really a hobby for me; it's more like an unbreakable habit. Maybe a vice. For some writers, this will change nothing, except to increase the avalanche of books, making it even harder to get noticed.

But, I have yet to see a discussion of how this tech might be used in marketing. I'm sure that conversation is already happening. When one door closes, a bottomless pit opens somewhere....

 

[CultureCitizen]

When will it happen? Now. Basic rule of thumb: if you are thinking about the future, it's probably already passed you by.

What will be the place of writers? Heck, writing for most people isn't profitable anyway. It's not really a hobby for me; it's more like an unbreakable habit. Maybe a vice. For some writers, this will change nothing, except to increase the avalanche of books, making it even harder to get noticed.

But, I have yet to see a discussion of how this tech might be used in marketing. I'm sure that conversation is already happening. When one door closes, a bottomless pit opens somewhere....

No, not yet, it can only produce blocks of 2,000 words and it also forgets anything beyond 2,000 words ... but maybe next year. GPT4 will have 1,000 more connections. If the hype is true, it will have as many connections as a human brain.
Marketing... there are already a number of apps written on top of gpt3 for those purposes: writing titles, copywriting etc.
The bright side: tv and streaming services will see an improvement in the scripts.

 

[Wayne Mack]

In the foreseeable future, I do not see AI creating a coherent novel or even a short story. I find that people typically overestimate the capabilities of computers and underestimate the complexity of human skillsets.

The understanding of how a novel is created is borderline mysticism. Go through a set of creative writing texts and one finds references to muses, the subconscious, left brain-right brain interaction. Ideas percolate up out of the unknown. It is easy to find descriptions of how plot lines should unfold, each of them different and each of them critiqued and rejected. In addition to plot, there are character arcs; the character must grow throughout the story, yet a great deal of popular characters show absolutely no growth. And in setting descriptions, there is the helpful advice of not having 'white room' settings and, when describing the setting, one should avoid 'purple prose.' Maybe one should strive for lavender environments.

Despite mankind's inability to define what a 'good' story entails or how it is created, there is a belief that the computer will discover the master pattern for stories. And to do this, AIs are 'trained' by ingesting the Internet, without any sort of qualifications or analysis. Instead of coherent patterns and story arcs, the result is an amalgamation of ideas and text that quickly shows inconsistencies. Furthermore, the generated text soon reflects some of the worst of ideas rather than being anything uplifting. Mankind does not know how to train an AI to write. It is even challenging to provide a curated list of 'good' writing--remember romance is by far the top selling genre. And there is certainly no known way to teach analysis of any provided text to identify what is good and what is bad. It is simply hoped that AI will make unknown leaps and bounds in analysis and be able to generate the ultimate novel.

I've been involved with computer software development for forty years and, in that time, AI has always been just a few years away. Some aspect of so-called AI will pop up and capture the general public's attention with associated gnashing of teeth over computers soon replacing us all. Then nothing significant happens and all is forgotten for another period of time. Revisit this thread in one year, and I predict the current set of AI programs will have been forgotten and gone the way of NFT hype.

 

[CultureCitizen]

In the foreseeable future, I do not see AI creating a coherent novel or even a short story. I find that people typically overestimate the capabilities of computers and underestimate the complexity of human skillsets.

The understanding of how a novel is created is borderline mysticism. Go through a set of creative writing texts and one finds references to muses, the subconscious, left brain-right brain interaction. Ideas percolate up out of the unknown. It is easy to find descriptions of how plot lines should unfold, each of them different and each of them critiqued and rejected. In addition to plot, there are character arcs; the character must grow throughout the story, yet a great deal of popular characters show absolutely no growth. And in setting descriptions, there is the helpful advice of not having 'white room' settings and, when describing the setting, one should avoid 'purple prose.' Maybe one should strive for lavender environments.

Despite mankind's inability to define what a 'good' story entails or how it is created, there is a belief that the computer will discover the master pattern for stories. And to do this, AIs are 'trained' by ingesting the Internet, without any sort of qualifications or analysis. Instead of coherent patterns and story arcs, the result is an amalgamation of ideas and text that quickly shows inconsistencies. Furthermore, the generated text soon reflects some of the worst of ideas rather than being anything uplifting. Mankind does not know how to train an AI to write. It is even challenging to provide a curated list of 'good' writing--remember romance is by far the top selling genre. And there is certainly no known way to teach analysis of any provided text to identify what is good and what is bad. It is simply hoped that AI will make unknown leaps and bounds in analysis and be able to generate the ultimate novel.

I've been involved with computer software development for forty years and, in that time, AI has always been just a few years away. Some aspect of so-called AI will pop up and capture the general public's attention with associated gnashing of teeth over computers soon replacing us all. Then nothing significant happens and all is forgotten for another period of time. Revisit this thread in one year, and I predict the current set of AI programs will have been forgotten and gone the way of NFT hype.

Novels can be written in several ways. But , since we are dealing with systems, a top-down structured approach will probably be the more feasible one for an AI : write the plot , create the characters, divide the plot into scenes, create dialogue for each scene, link the previous scene with the next one.

The current limitation of 2,000 tokens means they can't write anything which isn't a very short story.
AI has really advanced by strides in the past 5 years. I remember the first AI generated image ( made on Rembrandt's style) was made only 5 years ago, it required a lot of manual curation. Now it only takes a prompt, translation has improved a lot and it is amazing what GPT can translate ( as long as you feed 2,000 tokens per prompt). It can write code in the most popular languages and beat humans at any other strategy game.

If the hype is correct GPT4 will have 1,000 more connections than GPT3, and I am sure it will be as amazing as unsettling.

 

[sknox]

I think Microsoft's AI has larger limits already. I also think the 2k limit is more about server load than AI capabilities.

>overestimate the capabilities of computers
A couple of things about that. One, this isn't a computer. This is networks upon networks, which is a whole different ball game. The second thing is, I've been persuaded by Doug Hofstadter's argument ever since I first read it back around 1980. He argued that it won't be a matter of "true" artificial intelligence, but rather it will be when we humans simply start calling it that, treating it as intelligent. We claim all sorts of things are intelligent. The experts will continue to point out the distinctions, and the rest of society won't care.

 

[sknox]

As for the mystical nature of novels, I point to the avalanche of mediocre, pedestrian novels that have always found their way into print and have always sold at least somewhat. That's going to be the sweet spot for AI. Not timeless literature, just another entry into the potboiler stew. One need only browse through the staggering stacks at Amazon to see how very ordinary and formulaic are the majority of novels there.

If anything, I'd argue the short story market, such as still exists, is better proofed against AI. Why? Because magazines are very much a reflection of their editor--of, that is, a human being. AI will do best when writing to a more general and generic audience.

My worry isn't about AI, it's about humans. In particular, the scummy sort who will use armies of hourly workers to tell AI to churn out tidal waves of books that will generate a steady income for little work, complete with AI-generated covers, blurbs, and even reviews. This tidal wave will drown the honest authors who are hoping to make a living, or even just to get noticed.

Now I say it, I'm wondering if it might be time to go wide with my own books. Amazon may not be up to this task.

 

[Phyrebrat]

I doubt this will impact my enjoyment of writing or reading. It’s far more worrying regarding AI-art as far as I can see but the nuance of a book cannot be reproduced.

I think there’s a bit of Skyisfallingdown-ism with this AI malarkey. I think of it a bit like poor film adaptations of your favourite novels; a bad/inauthentic movie doesn’t negate the existence of the original. Same with this; real humans will still be writing even if robots are at it, too.

Tech often finds á different niche to its initial use so who knows where this might lead — archeological research improvements?

 

[msstice]

Whatever this AI is going to create is going to be pastiche (just like the art). At first glance it will be wow, but then the primary question on your mind will be, "From where is all this stolen from and glued together so cheaply?"

 

[sule]

I was listening to a podcast that was talking about AI-created books, and they brought up the US court case about the monkey selfies, in which the US Copyright court ruled that works created by a non-human are not copyrightable. I assume this means that an AI-written novel would also not be copyrightable, which means that, once published, it would be in the public domain and anyone could host it or use it for free - the publisher couldn't stop them and therefore wouldn't make as much money from it. (I am not a legal expert). Sure, a magazine or publisher could use an AI written story in place of one written by a human, but they wouldn't have exclusive rights to it because AI can't sign contracts.

I also don't think we need to be worried about an onslaught of AI-written stories/novels ruining writing. There is a LOT of writing floating around for free on the internet and, as far as I can tell, it hasn't tanked the public's desire for stories. It is possible that AI changes how people consume stories, but for now I think we're safe from a complete AI takeover of the literary world.

In conclusion: AI novels won't be much of a factor in publishing because publishers won't be able to make as much money from them as from a human novel, and a glut of AI stories (available for free) isn't likely to tank the traditional market for stories.

 

[Steve Harrison]

One thing that occurred to me is that to create a novel or non-fiction book, AI has to draw upon all works of fiction online, which would leave the publisher open to a class action plagiarism lawsuit, particularly if a bestseller.

If that's the case (pardon the pun), there's hope for us yet!

 

[CultureCitizen]

One thing that occurred to me is that to create a novel or non-fiction book, AI has to draw upon all works of fiction online, which would leave the publisher open to a class action plagiarism lawsuit, particularly if a bestseller.

If that's the case (pardon the pun), there's hope for us yet!

Not really , it can read all the novels that are not subject to copyright ( anything prior to 1923).
The models learn patterns , and the relation between words and characters , they actually build abstractions.
That means that if some writer came up with a novel writing style it will not be able to create a similar output.
Science fiction novels "might" be a bit cheesy or unimaginative.

 

[Steve Harrison]

Not really , it can read all the novels that are not subject to copyright ( anything prior to 1923).
The models learn patterns , and the relation between words and characters , they actually build abstractions.
That means that if some writer came up with a novel writing style it will not be able to create a similar output.
Science fiction novels "might" be a bit cheesy or unimaginative.

It was worth a try!

 

[OuttaInc]

According to Clarkesworld, AI-written submissions are skyrocketing. Happily they’re easy to detect (for now) but their overwhelming numbers stress an already overworked system. In short, it’s not the quality of AI writing that’s the problem, it’s the quantity.

A Concerning Trend – Neil Clarke

An AI that writes bad science fiction... surely there’s a story in there.