question on the use of ai to help with writing

[deathmagicuser]

So I recently started using an ai chatbot service that uses different scenarios, and I wanted to turn the scenarios I go through into stories would that be something I could do so long as I disclaimer that it was made with the help of ai

 

[BronzePlaceWriter]

So I recently started using an ai chatbot service that uses different scenarios, and I wanted to turn the scenarios I go through into stories would that be something I could do so long as I disclaimer that it was made with the help of ai

Click to shrink...

I don't think we have any rules against it. Some people won't like it and won't follow your threads over doing it, though.

 

[deathmagicuser]

I don't think we have any rules against it. Some people won't like it and won't follow your threads over doing it, though.

Click to shrink...

I just want to make sure with the mods first ya know

 

[riotous living]

I just want to make sure with the mods first ya know

Click to shrink...

They've talked about it before in this forum

 

[deathmagicuser]

Welp if the mods don't like it they'll let me know by messaging me and taking it down

 

[Renko]

People are mostly fine with AI if it's for grammar and spelling though.

 

[deathmagicuser]

People are mostly fine with AI if it's for grammar and spelling though.

Click to shrink...

the ai will be doing most of the writing it is a chatbot after all basically I open the scenario read the opening message type in a response the ai sends another message based on both my responses and the contextual information provided with the scenario I will of course be editing it slightly to better fit a story format

 

[NumberSix]

So you are roleplaying with an A.I. and want to turn it into a story after editing it and fixing some A.I. errors?
I mean, I really don't see why anyone should have a peoblem with that, and last I checked A.I. outputs can't be copyrighted, especially if you are running an open source model locally.

Personally I've tried to use A.I. as an assistant, like say asking it to give me a sunonim for excited pouting in a sexual context.


Is it going to be a lewd story?

 

[JadeKaiser]

Technically speaking, you don't even have to state that it was made with the help of AI, as there's no rule against it. Though, using AI for your writing without saying so is generally considered to be "a dick move" by a lot of people. Anything that is even partially AI-generated tends to be considered as "slop" by a lot of people, and it generally comes out sooner or later. Pretty much any level of AI use in the writing process tends to show through on some level or other, at least to some people, and when the writer tries to hide it, it creates a lot more backlash than if they were open about it. Even when people are open about it though, there will be people who complain and more people who simply choose not to read because of it.

If you want to avoid that, but still want to use the stories from those AI RP sessions because you just genuinely like how the story turned out and want to share it as one, I would actually suggest something else. Something that I haven't seen anybody do before: do your RP session, using the AI like a GM. Save it. Then, using none of the AI-generated text directly, write up the story that evolved from it as a separate document, by yourself in all your own words. Essentially, using the RP session as a storyboard or outline.

I feel like there's still probably a chance it could create bad writing habits or something, particularly if you clung too closely to the plot as generated by the AI instead of diverging when you felt you had better ideas, but we probably can't know until someone tries it.

 

[SmokeRichards]

I feel like there's still probably a chance it could create bad writing habits or something, particularly if you clung too closely to the plot as generated by the AI instead of diverging when you felt you had better ideas, but we probably can't know until someone tries it.

Click to shrink...

It might be the ego, but I consider myself a very good writer.

In my experience, AI tend to be very formulaic, and have issues with coherency. They are unable to keep track of enough details to tie up all the plot threads in a way that is compelling. The formulaic nature is the more pressing issue, however. An AI is probably going to make something that is, at least on surface, on par or better than the average fandom work, but because of the way AI functions, they are more likely to follow trends, either on a fandom level, or on a more overarching literature level when it comes to writing as a whole across society. AI are going to mix and match things that are already in fashion, and while you can guide them away from this tendency, it is like attempting to hold back the sea. The water is always going to return. I have tried to use AI for brainstorming, and always found the output frustratingly recursive. By the time you are attempting to guide it into what you want, you already knew what you needed in the broad outlines.

A writer needs to have a coherent vision for what they want to achieve with AI if they want to make something good, but by that same token, if you have that coherent vision, you don't need the AI, even if the AI can do the writing faster.

The one place I think AI has a place is going to be prose. AI prose is not going to be winning awards any time soon, but compared to the average fandom writer, what AI can do is very good. A writer could learn a lot from have AI rewrite sections of story and comparing it to the original to see what is being changed and why it works better.

TLDR: It's a tool, more powerful than most, but a predictive engine is never going to be able to narrow down on what makes a story unique. Predictive engines are all about finding averages, which is why they work, and why they will never excel.

 

[NumberSix]

Technically speaking, you don't even have to state that it was made with the help of AI, as there's no rule against it. Though, using AI for your writing without saying so is generally considered to be "a dick move" by a lot of people. Anything that is even partially AI-generated tends to be considered as "slop" by a lot of people, and it generally comes out sooner or later. Pretty much any level of AI use in the writing process tends to show through on some level or other, at least to some people, and when the writer tries to hide it, it creates a lot more backlash than if they were open about it. Even when people are open about it though, there will be people who complain and more people who simply choose not to read because of it.

Click to shrink...

If somebody is using it as a glorified search engine/thesaurus/dictionary/style assistant/proof reader I don't think it is a dick move.

How would I rephrase xyz in the style of or what is the idiom/potential rephrase for something.

That would not be rewriting but using the A.I. for research purposes.

It won't be any different than clipping phrases you like and reusing them or using Google or grammarly.

 

[JadeKaiser]

If somebody is using it as a glorified search engine/thesaurus/dictionary/style assistant/proof reader I don't think it is a dick move.

How would I rephrase xyz in the style of or what is the idiom/potential rephrase for something.

That would not be rewriting but using the A.I. for research purposes.

It won't be any different than clipping phrases you like and reusing them or using Google or grammarly.

Click to shrink...

<Shrugs

I wasn't stating an opinion, just explaining the sort of reactions one might expect. People don't like it when someone uses AI and appears to be trying to hide it. What you're describing isn't really relevant to the question that was asked.

Also, about half of what you're describing is just a really bad idea that will screw you over because AI/LLMs aren't made for that, and will confidently tell you something that's completely wrong. LLMs aren't made to answer questions or be used as a search engine, they're made to fit words together in a way that looks nice. If you ask it a question, it'll give you an "answer," but the actual correct answer to your question will have at most, tertiary relevance to what that "answer" will be. Just look at all the times people have clipped screenshots of AI telling them that it's safe to drink bleach and stuff like that, for example.

 

[Pure Unadulterated Ego]

Also, about half of what you're describing is just a really bad idea that will screw you over because AI/LLMs aren't made for that, and will confidently tell you something that's completely wrong. LLMs aren't made to answer questions or be used as a search engine, they're made to fit words together in a way that looks nice. If you ask it a question, it'll give you an "answer," but the actual correct answer to your question will have at most, tertiary relevance to what that "answer" will be. Just look at all the times people have clipped screenshots of AI telling them that it's safe to drink bleach and stuff like that, for example.

Click to shrink...

Honestly, I don't understand what you're talking about. LLMs excel at rephrasing things or asking general questions like what idioms fit in a certain situation. It's actually one of the most basic, boilerplate things they are used for (part of the reason they are so used for mails).

You shouldn't trust them with facts, which is what you gave as an example and I agree, but it's ridiculous to take that as an example of not using them as the tool they are designed as. You should always be the barrier/director that stands in place and says if it's good or not.

In my experience, AI tend to be very formulaic, and have issues with coherency. They are unable to keep track of enough details to tie up all the plot threads in a way that is compelling. The formulaic nature is the more pressing issue, however. An AI is probably going to make something that is, at least on surface, on par or better than the average fandom work, but because of the way AI functions, they are more likely to follow trends, either on a fandom level, or on a more overarching literature level when it comes to writing as a whole across society. AI are going to mix and match things that are already in fashion, and while you can guide them away from this tendency, it is like attempting to hold back the sea.

Click to shrink...

On a more general sense... AI is going to be shit at giving you an actual finished story (at least with current models and in my opinion). That said, this doesn't really mean you shouldn't use it as a tool for one. Having issues with coherency is fine from time to time, you should be the one checking that particular aspect.

Some of the things mentioned in this thread in general generalize so much that are borderline useless/inaccurate. You need to take into account the model you're using (huge difference between opensource 7B model, GPT or Claude/Gemini) and how you are actually using the thing (like in image gen, you can use a good model to gen slop). The part of mixing and matching things in fashion is just wrong though. Maybe if you just used "write a Naruto fanfic about an Uzumaki orphan" rather than actually getting good at using a new tool.

You can have it write a first draft following certain guidelines/style, and then you edit. Or have it write the outline that you edit and then you go for a first draft. Or give it a book on writing, summarize it as general things to follow and then use it to nitpick your own writing to death. Or brainstorm things for worldbuilding or so you get unstuck on a turn of phrase... and so on. There are plenty of people using it in a ton of different ways that aren't just Grammarly or posting the first answer the model gives.

I have tried to use AI for brainstorming, and always found the output frustratingly recursive. By the time you are attempting to guide it into what you want, you already knew what you needed in the broad outlines.

A writer needs to have a coherent vision for what they want to achieve with AI if they want to make something good, but by that same token, if you have that coherent vision, you don't need the AI, even if the AI can do the writing faster.

Click to shrink...

Not really true in my experience, both in writing and elsewhere. Sometimes AI is pretty good at being an enhanced rubber duck and it comes with pretty cool ideas from time to time. Especially because the big models are much more capable in different subjects than a single human can. What you create can and will change depending on your process. It has even happened with image gen when the AI does something slightly different and I have to admit it does look better than what I was thinking of.

 

[NumberSix]

Honestly, I don't understand what you're talking about. LLMs excel at rephrasing things or asking general questions like what idioms fit in a certain situation. It's actually one of the most basic, boilerplate things they are used for (part of the reason they are so used for mails).

You shouldn't trust them with facts, which is what you gave as an example and I agree, but it's ridiculous to take that as an example of not using them as the tool they are designed as. You should always be the barrier/director that stands in place and says if it's good or not.

On a more general sense... AI is going to be shit at giving you an actual finished story (at least with current models and in my opinion). That said, this doesn't really mean you shouldn't use it as a tool for one. Having issues with coherency is fine from time to time, you should be the one checking that particular aspect.

Some of the things mentioned in this thread in general generalize so much that are borderline useless/inaccurate. You need to take into account the model you're using (huge difference between opensource 7B model, GPT or Claude/Gemini) and how you are actually using the thing (like in image gen, you can use a good model to gen slop). The part of mixing and matching things in fashion is just wrong though. Maybe if you just used "write a Naruto fanfic about an Uzumaki orphan" rather than actually getting good at using a new tool.

You can have it write a first draft following certain guidelines/style, and then you edit. Or have it write the outline that you edit and then you go for a first draft. Or give it a book on writing, summarize it as general things to follow and then use it to nitpick your own writing to death. Or brainstorm things for worldbuilding or so you get unstuck on a turn of phrase... and so on. There are plenty of people using it in a ton of different ways that aren't just Grammarly or posting the first answer the model gives.

Not really true in my experience, both in writing and elsewhere. Sometimes AI is pretty good at being an enhanced rubber duck and it comes with pretty cool ideas from time to time. Especially because the big models are much more capable in different subjects than a single human can. What you create can and will change depending on your process. It has even happened with image gen when the AI does something slightly different and I have to admit it does look better than what I was thinking of.

Click to shrink...

Yeah, and most models have some degree of error correction built in and are capable of calling various APIs to validate real data and the models are refactored and the date put into them cherry picked.

The fact that some result is generated pseudorandomly doesn't mean that there aren't any guardrails and quality gates.

And if the answers are nondeterministic the results needen't be either.

There really isn't a perfect way to phrase a statement or describe some fictional prop.

It is not like a fanfic's technobabel for example needs to include working orbital mechanics for instance, Arthur Clarke might have done that for short stories but most writes are not him.

And the A.I. can give me a few ways to rephrase "...she stared at him with bedroom eyes..." To make the peose sound more interesting then I will use it from time to time.

Not being a native speaker it can just give me more ideas how to do stuff, not that I use the suggestions verbatim, anyways.

Here is a little example.

Spoiler: screenshots

Prompt:

Rephrase the below to sound more formal, generate five alternative rephraisals: And a lot of models muse vsrious forms of checking ansnpeuning to improve results and narrow down proper responses.
Fix spelling errors and provide rationale for the generated outputs.

And a lot of models muse vsrious forms of checking ansnpeuning to improve results and narrow down proper responses.
Response:

Here are five formal rephrasings of your sentence, along with corrections for spelling errors and a rationale for each output:
Original (with corrections):

"And a lot of models use various forms of checking and pruning to improve results and narrow down proper responses."
Rephrased Versions:

"Many models employ various forms of verification and pruning to enhance results and refine response accuracy."

Rationale: Replaces informal terms ("a lot of" → "many," "use" → "employ") and clarifies intent ("narrow down proper responses" → "refine response accuracy").

"Numerous models utilize different validation and pruning techniques to optimize outcomes and ensure precise responses."

Rationale: Enhances formality ("a lot of" → "numerous," "improve results" → "optimize outcomes") and specifies "validation" for clarity.

"A significant number of models incorporate various checking and pruning mechanisms to boost performance and filter responses effectively."

Rationale: More precise phrasing ("a lot of" → "a significant number of") and stronger verbs ("improve" → "boost," "narrow down" → "filter").

"Multiple models leverage diverse verification and pruning strategies to enhance output quality and restrict responses to relevant ones."

Rationale: Uses "leverage" for sophistication, "diverse" for variety, and "restrict responses to relevant ones" for precision.

"Various models implement systematic checking and pruning methods to improve accuracy and select the most appropriate responses."

Rationale: "Systematic" adds rigor, while "select the most appropriate responses" is more formal than "narrow down proper responses."

Spelling Corrections & Rationale:

"muse" → "use" (Typo)

"vsrious" → "various" (Typo)

"ansnpeuning" → "and pruning" (Miskeyed letters)

These revisions improve professionalism, eliminate informal phrasing, and ensure technical accuracy. Let me know if you'd like further refinements!
New chat
AI-generated, for reference only

Click to shrink...

I think it is pretty good, even better at detecting typos and turning them into something actually workable, too.
Sure bests grammarly in that regard.

Edit: Oh, yeah, forgot to mention, this is a lot like the typewriter vs. Dedicated word processor vs. PC with ancient dos stuff like WordStar vs. actually modern office software.

I am pretty sure at least one scifi writer bitched about muh computors not being the same as his old typewriter.
Cough-cough, I think it was Harlan Ellison, cough-cough.

And a bunch of scifi writers still had to use this ancient thing called WordStar.

And among a particular subset of it people there is always the old Vi vs. EMACS vs. Nano omit:
What often gets those old farts to bury the hatchet is hatered for actual modern IDEs and advanced editors like VS code.
Also, Vim is best, the other CLI/TUI ones are crap, there is a reason the first hit for EMACS is "how do I exit it", 2 key combos beats 3.
Seriously, A.I. can be just another tool and should be treated as just another tool.

But do not let it do your thinking for you.

 

[Fellgar]

So I recently started using an ai chatbot service that uses different scenarios, and I wanted to turn the scenarios I go through into stories would that be something I could do so long as I disclaimer that it was made with the help of ai

Click to shrink...

Which chatbot are people using? Are they any good?

 

[deathmagicuser]

Which chatbot are people using? Are they any good?

Click to shrink...

I was using JanitorAI and it can be a little wonky sometimes but the bots themselves are pretty good most of the time

 

[Fellgar]

I was using JanitorAI and it can be a little wonky sometimes but the bots themselves are pretty good most of the time

Click to shrink...

Huh. So I take it you select the character then and just have dialogue? Not sure how else that particular ai can help with writing. Thabks!

 

[deathmagicuser]

Huh. So I take it you select the character then and just have dialogue? Not sure how else that particular ai can help with writing. Thabks!

Click to shrink...

Yeah my stories were literally just transcripts of my conversation with the bots

 

[SuperPulp2789]

Huh. So I take it you select the character then and just have dialogue? Not sure how else that particular ai can help with writing. Thabks!

Click to shrink...

Helps more properly visualize things, I guess

Part of the problem for many creatives and would-be-creatives is getting the ball rolling and actually having to redo things again and again

To make it right, if the creators and their editors even actually bother

Writer's Block and over perfectionism makes things harder than they need to be

It's why I am of the belief that AI should be for mainly making a rough sketch and helping the actual human do the redo and finishing touches

The reason why so many people will STILL do traditional/digital art? Because there are limits to prompting and because even when technology advanced, people still used the old ways

Hell, from what I know, even the Amish adopt some modern-ish tech so long as they can reproduce it themselves(these guys build their own houses, not just hire the rest to build it for them) and avoid being over dependent

Trust me, even as AI gets better, it will STILL feel off, same as hiring somebody else, in part it doesn't conform to your exact vision

 

[Doctor Blood]

Edit: Oh, yeah, forgot to mention, this is a lot like the typewriter vs. Dedicated word processor vs. PC with ancient dos stuff like WordStar vs. actually modern office software.

Click to shrink...

An IBM Selectric, MicroPro Wordstar, Corel WordPerfect, and Google Docs are different in degree, while LLMs are a difference in kind.

Using an LLM is like delegating work to a mentally unstable idiot-savant; give it the correct tasks, structured the right way, and it can do miracles; give it the wrong task or the wrong prompt and you'll get garbage. Sometimes it gives you nonsense no matter what you try. The onus is on you, the user, to parse and validate the output.

The fundamental problem with using an LLM to do most of your writing is that you've relegated yourself to the role of a commissioner and editor, and the kind of people that outsource their writing to an LLM rarely have the necessary skills and sense of taste to edit the final work into something acceptable.

As for 'actually modern' office software... there's a reason there were so many holdouts for tools like Wordstar and WordPerfect - they have features that GDocs and MS Word still lack. A Writer's Word Processor by Robert J. Sawyer is worth a read.

Also, Vim is best, the other CLI/TUI ones are crap, there is a reason the first hit for EMACS is "how do I exit it", 2 key combos beats 3.

Click to shrink...

Every word you type belies your ignorance. People joke that Emacs is an operating system, but this is more true than they realize; it's a third-rate implementation of a lisp machine. However - even an incomplete, hacky reproduction of a lisp machine stands head and shoulders above 'modern' editors like VS Code, nevermind toys like vi and nano. Emacs looks strange and intimidating at first, but that's because it was originally developed on a PDP-10 and predates modern GUI terminology, CUA shortcuts, and the IBM keyboard.

Why Emacs? Because it's the advanced, extensible, customizable, self-documenting editor. Aside from the core lisp interpreter written in C, the entire environment is written in elisp - which means that you can use the repl to modify every function in Emacs in real time. You can incrementally and interactively build tools to do anything you want, or change how any part of Emacs works while it's running.

On top of that, packages and built-in tools like Calc, Eshell, Hyperbole, Helm, TRAMP, Magit, SLIME, Org Mode, AUCTeX, ERC, elfeed, and Gnus make Emacs by far the most powerful and flexible computing environment - and all of these tools can be used through a unified environment, and any of their components can be made to interface with each other. See also Why Emacs itself is my "favorite Emacs package."

. . . I use Emacs, which might be thought of as a thermonuclear word processor. . . . Emacs outshines all other editing software in approximately the same way that the noonday sun does the stars. It is not just bigger and brighter; it simply makes everything else vanish."
-- In the Beginning Was the Command Line by Neal Stephenson

Click to shrink...

Emacs is a descendent of a different evolutionary branch of computer interaction, a window into another world - a world where computers don't suck so much.

Now get off my lawn!

(Of course, for real men - Ed is the standard text editor.)
(Bonus: Interview with an Emacs Enthusiast)

 

[brokenBillairds]

Please just correctly tag your story with 'ai assisted' or something if you do

 

[Lady Naga]

EMACS isn't AI it's just a word processor from 1976.

 

[Zoroastrian]

What would count as "AI Assisted" though? If I used AI to make dialog and copy pasted that into my story that's obviously made toh AI. But what if I were to give the chat bot a character and parameters to pretend to have a certain personality and backstory, then ask it questions to figure out how that character would respond? What if I asked the AI to create backstory for a character that I created? What if I asked it to create a character wholesale but wrote them myself? What if I used AI to create a custom random encounter table and used that to figure out events (I have done this)?

There are many, many types of "AI Assisted" content one could feasibly have. Where to draw the line? Is writing a story and using an AI image "AI Assisted" (I have done this too)?

 

[dasstan]

What would count as "AI Assisted" though? If I used AI to make dialog and copy pasted that into my story that's obviously made toh AI. But what if I were to give the chat bot a character and parameters to pretend to have a certain personality and backstory, then ask it questions to figure out how that character would respond? What if I asked the AI to create backstory for a character that I created? What if I asked it to create a character wholesale but wrote them myself? What if I used AI to create a custom random encounter table and used that to figure out events (I have done this)?

There are many, many types of "AI Assisted" content one could feasibly have. Where to draw the line? Is writing a story and using an AI image "AI Assisted" (I have done this too)?

Click to shrink...

The people complaining/worrying about AI assistance with writing just don't want to read stories written by AI.

Submit a draft of your own writing and use the AI to highlight issues and suggest how it can be improved for quality? Great. Go for it. It's still your words. Even if you do implement the changes suggested by the AI, the meat of the story will be largely your own original writing and ideas.

Submit a prompt and let the AI do most of the work? I'd count that as the kind of AI assistance that would cause issues for the people that don't want to read it.

The end product should be crafted by human hands.

 

[Zoroastrian]

The people complaining/worrying about AI assistance with writing just don't want to read stories written by AI.

Submit a draft of your own writing and use the AI to highlight issues and suggest how it can be improved for quality? Great. Go for it. It's still your words. Even if you do implement the changes suggested by the AI, the meat of the story will be largely your own original writing and ideas.

Submit a prompt and let the AI do most of the work? I'd count that as the kind of AI assistance that would cause issues for the people that don't want to read it.

The end product should be crafted by human hands.

Click to shrink...

But that's not AI assistance, that's just written by AI. At least some of the people in this thread want any and all AI assistance to be tagged, but the question is what that even means. Where do you draw the line? Obviously just pasting the AI into a forum post is way more than just assistance, but if the AI tells you what to do and you then write it in your own words is that AI assistance, and if it is could you even tell?

I can write. I'm even good at it. I also frankly, give no shits if people want to use AI to write for them as long as they aren't trying to hide it. But even if they are hiding it it's really quite hard to tell what's AI and what's just a bad writer. The end result, though, is the same. Just ignore it and move on