ChatGPT For World Building?

[_Michael_]

I've been using ChatGPT for awhile to sort of fill in some of the details and give me blocks of expository information that I then edit and include. Anyone else do this? I know ChatGPT isn't perfect, but with proper supervision, it's pretty useful, and I've even used it to flesh out some basic rules for an RPG system based on 3.5e + FNFF.

Caveats: You will occasionally have to get it back on format if you're using it to generate paragraphs in a formatted fashion, and it's memory isn't infinite, so it's best to take notes externally as you go so you don't lose stuff after a week or so of talking to it.

 

[writer_megan]

I've been using ChatGPT for awhile to sort of fill in some of the details and give me blocks of expository information that I then edit and include. Anyone else do this? I know ChatGPT isn't perfect, but with proper supervision, it's pretty useful, and I've even used it to flesh out some basic rules for an RPG system based on 3.5e + FNFF.

Caveats: You will occasionally have to get it back on format if you're using it to generate paragraphs in a formatted fashion, and it's memory isn't infinite, so it's best to take notes externally as you go so you don't lose stuff after a week or so of talking to it.

I’ve used ChatGPT before for some idea generation. Sometimes I don’t end up using anything generated but then make changes or at least it gets the ball rolling.

 

[_Michael_]

The more you play with it, the more refined you can get--to a point. It's definitely more of a seed generator than anything else. The rpg stuff I was using it for was strictly for sketching the outline of the system--I wouldn't trust it with the math portions. But it did help me generate a really cool set of sister planets, one with two moons in a co-planar moons (meaning one moon orbits around the northern hemisphere, and one orbits the southern, and they never cross) in a trojan configuration at a near-perfect 4:1 resonance (the smaller moon orbits 4 times for every 1 orbit of the larger--resulting in an extremely stable orbit). 438 solar day year with a 3 day correction period at the end called the Lunarion and a single correction day at the end of each 4 week, 7-day month called Albdas.

Expository text it has more trouble with, but it's still useful to give you a few good sentences like seeds to expand on.

Plus, I think they just updated it to be able to search anything on the internet past 2021 or whenever the cutoff was.

 

[writer_megan]

The more you play with it, the more refined you can get--to a point. It's definitely more of a seed generator than anything else. The rpg stuff I was using it for was strictly for sketching the outline of the system--I wouldn't trust it with the math portions. But it did help me generate a really cool set of sister planets, one with two moons in a co-planar moons (meaning one moon orbits around the northern hemisphere, and one orbits the southern, and they never cross) in a trojan configuration at a near-perfect 4:1 resonance (the smaller moon orbits 4 times for every 1 orbit of the larger--resulting in an extremely stable orbit). 438 solar day year with a 3 day correction period at the end called the Lunarion and a single correction day at the end of each 4 week, 7-day month called Albdas.

Expository text it has more trouble with, but it's still useful to give you a few good sentences like seeds to expand on.

Plus, I think they just updated it to be able to search anything on the internet past 2021 or whenever the cutoff was.

Yeah I feel like it’s super useful for fact checking or research and making sure the world is functionally believable. I just wonder if anything ever gets published like is there a copyright thing or how would anyone even be able to tell if an idea or sentence came from chat gpt

 

[_Michael_]

I don't think there is, especially if you're changing text around and putting a lot of work into it. I'm up to like 50 pages or so, and right now, I'm working my way through the list of deities in the overaching pantheon I call the Golden Pantheon.

Fact checking your own stuff can be problematic--best to take notes. Sometimes it will repeat a name from a previous query. Generally, it's pretty reliable.

 

[A. E. Lowan]

Yeah I feel like it’s super useful for fact checking or research and making sure the world is functionally believable. I just wonder if anything ever gets published like is there a copyright thing or how would anyone even be able to tell if an idea or sentence came from chat gpt

Yes, and the Writer's Guild of America just went to strike over this and related matters. And yes, sentences and phrases (not ideas, you can't copywrite ideas) are as recognizable as the facial features of movie stars hidden in AI artwork. You can especially see them when they're yours.

I wouldn't see anything wrong with using the thing to spitball and brainstorm, but I do recommend a program like OneNote to keep track of your growing store of lore. It's searchable and versatile, and my team has used it for over 15 years to keep track of our various projects, including a long-running Urban Fantasy series, and it's kept up.

 

[Devor]

I just wonder if anything ever gets published like is there a copyright thing or how would anyone even be able to tell if an idea or sentence came from chat gpt

So far, courts have ruled that human authorship is an essential part of copyright, and that AI-generated content has no owner. It dates back to a case where a chimp took a photo, and the photographer who set it all up lost any rights to it.

But then again, AI-generated text still needs a lot of work to be publishable, and that work is copyrighted. If there's an AI image, and you run it through filters and make edits, those changes do get copyrighted. And it's hard for anyone to know what was or wasn't part of the original piece which may not even have been saved. So the law so far currently favors the AI user over the company that made the algorithm, over anyone who would repurpose the generated content, and over anyone whose work may be subtly part of the algorithm.

 

[Finchbearer]

Chat GTP doesn’t fact check - so be aware of this. I know that students are being caught out when asking chat GPT for references, it simple gives fake ones that look very convincing. So it will throw lots of information at you - just like the internet does - but may not be actual facts. Search for multiple sources to be (slightly more) sure.

 

[Devor]

I've done my share of tinkering with it. I found it helpful initially, but I'd already completed most of the planning work it might've helped out with. I'd really love to use AI art and put together some kind of tabletop game, but there's only so many projects I can pursue. Maybe some day though.

It's definitely useful for anything that's straightforward and fact-based (if you can spot the bullshit), but it's not there yet with creative writing, even with tons of guidance. In terms of creative writing, I find it's main use to be prompts like "Tell me twenty details you might find walking through a medieval fantasy tavern." Then grabbing a few of them for your own description.

 

[pmmg]

I am sure AI can write stuff. Why do we need you?

 

[A. E. Lowan]

I am sure AI can write stuff. Why do we need you?

And hence the strike.

 

[_Michael_]

AI writes stuff poorly. If you ask it to generate lore without parameters, you're gonna get a mess. Fact checking is of limited use in fantasy writing and world-building--you're mostly inputting your own "facts" and hoping it remembers them long enough to generate usable results.

It would be a lot harder to prove AI generated content in text (especially if you're changing things and editing as you go) than it is with AI images (which can be run through free AI detectors). I use AI artwork right now as a placeholder and to show the general ideas I want represented prior to the final draft.

As for AI replacing writers, I'm not too concerned. The strike I don't care about because it's been a long time coming, and should have been the end of the SAG. No one needs Hollywood and no one cares about their opinions outside of the silver screen. However, they've become so arrogant that they think that anything they say or do is of utmost cultural importance when the opposite is true. Ai being used as a tool to generate new seeds of ideas is A-okay by me because anything has got to be better than Yet Another Scary Movie sequel.

For world building, it's much the same. It will also critique your stuff as well, if you enter parameters like, "Compare this to the work of RA Salvatore." Or if you ask it to consider other franchises while generating your idea, like, "For my idea [followed by the explanation], I'd like you to also consider the Wheel of Time series in your assessment."

 

[pmmg]

And what will you do when it doesn't write it poorly?

I forgot about AI. I just completed Nano Yay, me. Where do I get my badge?

 

[Finchbearer]

I do a lot of fact-checking for my fantasy worlds - what the cloistered life of 13th century novices was like, what plants created certain coloured dyes and how they were processed, terminology for items of clothing, horsemanship, anything ship related…the list goes on. You don’t do fact-checking for your fantasy writing?

 

[_Michael_]

I don't. I use startpage for searching historical facts after getting key terms from the AI. If I ask it what the cloistered life of 13th century novices was like, whatever answer it gives me I then use to inform my search engine queries. Otherwise, it will tell you that the wimple was a great fashion piece in one sentence, and then tell you that the wimple was a product of ancient Egyptian design in the next. You use it to get terms and then use those terms to further explore in other sources. Relying upon the AI entirely is a fool's errand. Yes, it's getting better, and yes, it was opened up to be able to search the internet, but it will still make up stuff if you permit it to.

Hence why I use it to generate expositional text for filler, based on carefully inputed parameters. I may ask it what the name of a term is, but I double check it to make sure it is what the AI says it is.

It's a very useful tool, to be sure, but one that, like your average six year old child, requires constant supervision to keep it from running off on a tangent. So as far as "fact-checking" goes, i don't really use it as such, if I need to know something beyond asking for specific terms or conditions of life. Same goes for the3 statistical side of RPG system building. I used it to sketch out the rough outline of my ideas, but I don't ask it to figure out the percentages or to weigh tables, and if I do, I double-check it to be sure. I've caught it a couple times recycling names in improper places from what I've inputted (like re-using the name of a god when I was doing the deities).

 

[Finchbearer]

I meant in general - you said; Fact checking is of limited use in fantasy writing and world-building--you're mostly inputting your own "facts" and hoping it remembers them long enough to generate usable results.

Sure you make your own lore up when writing fantasy, but I mean the kind of facts you need to fact check, you sounded as if you didn’t do that for your writing was all I got from it.

 

[Demesnedenoir]

AI is a toy to me. I would spend more time dicking with ChatGPT and correcting it's crap than actually writing to make it work. Even for pure world-building, it sucks for me. Maybe if I was working from scratch it would be more useful. The AI Art is a better toy for inspiration. Although I didn't use anything it produced, I tinkered around with a variety of thoughts on flags and what I generated pointed me in curious directions that led to a flag description that didn't even make it into the book... Fun, but in the end, still wasted more time than I should have, LMAO.

 

[Ban]

I made a thread a while ago wherein I tested Bing Chat's worldbuilding abilities, but I have no intention or interest of using it myself for any of my works. It can be a good tool for beginning writers or tabletop rpg GM's, but for the long-term it is better to train your own worldbuilding skills. A good launchpad, in the same manner as those online generators on Donjon.bin.sh for example are.

 

[A. E. Lowan]

AI is a toy to me. I would spend more time dicking with ChatGPT and correcting it's crap than actually writing to make it work. Even for pure world-building, it sucks for me. Maybe if I was working from scratch it would be more useful. The AI Art is a better toy for inspiration. Although I didn't use anything it produced, I tinkered around with a variety of thoughts on flags and what I generated pointed me in curious directions that led to a flag description that didn't even make it into the book... Fun, but in the end, still wasted more time than I should have, LMAO.

Sounds like the codependent relationship I have with Pinterest.

 

[Demesnedenoir]

Fortunately, I have an antagonistic relationship with Pinterest. It's a mutual thing.

Sounds like the codependent relationship I have with Pinterest.