ChatGPT For World Building?

[_Michael_]

They must have updated it because it's craptacular now. I ask it for expository descriptions based on (NOT plagiarizing) the text block I give as an example), and it just regurgitates exactly what I wrote. I've tested this now on new chats, and it's the same. It's like ChatGPT suffered a stroke sometime last week, and now, is nothing but an exercise in frustration.

This is to say nothing of the Sam Altman firing, near-rehiring, and sudden poaching by Microsoft.

Whatever they did, they borked it entirely. Now, I have to fight through twenty minutes of arguing with a dumbass (because that's what it has become overnight) to try to wrestle out usable information. Out of EIGHT paragraphs of plagiarized and mixed up bullshit, I got a three sentence paragraph. Wtf.

Are there any other ChatGPT sites that work just as good as it used to? Gotta be an alternative.

 

[pmmg]

How about being a writer instead.

 

[_Michael_]

Can't I do both? lol For me, it's more like someone I can bounce ideas off of. The useful stuff I get I remix and then feed it back for critiques. Paragraph after paragraph.

 

[Fyri]

With hesitation, I will mention character.ai

It is a website created for people to essentially get to make and talk to bots that resemble people (fictional or real). There are a few "Creative Writing" bots on there that you can use to bounce ideas off of.

I think AI is a useful tool, like a thesaurus, but A.E. Lowan is right. If you want a trustworthy answer, you're gonna need to get human feedback—and not just any human. A human that has experience and can be credible in their guess for your creative purpose. If you use AI, I'd say to just... proceed with caution?

Like for me, I once needed to know how lava and ice react to each other in a specific condition for a fight scene. I could have asked an AI (didn't exist at the time) but the answer may have been completely wrong. My critique partner friend read the scene later and questioned the interaction of elements, but I had watched videos on youtube that mimicked the exact situation and also combed through a lot of articles about how lava and ice work together (I had to dodge so many articles about making obsidian in Minecraft DX).

If someone questions your decisions in your story, do you want to refer back to a chat bot that made things up based on random info it pulled from online, or do you want to be able to cite an article written by someone that actually studied the stuff specifically?

Point is, these AI sites are great tools, but, for the most part, are making everything up. They are chat bots. They don't have real experience or expertise about the world, besides what is already online, and they use that information only for conversation purposes—not with the ability to ensure they are right or logical.

If you're wanting to bounce your ideas off of an AI bot, you'll have to keep in mind that you'll need to do more work yourself to ensure the answer from the bot is something you agree with / that makes sense. Here, you are already noticing the frustrations of working with Language-based Artificial Intelligence:

Now, I have to fight through twenty minutes of arguing with a dumbass (because that's what it has become overnight) to try to wrestle out usable information. Out of EIGHT paragraphs of plagiarized and mixed up bullshit, I got a three sentence paragraph. Wtf.

It's not always reliable for it's basic function, let alone crafting something made-up that other humans will find plausible. This is not the first AI site to have this complaint.

However, we are writing fantasy. There is a magic in versimilitude.

 

[_Michael_]

Spot-on, my friend. It is indeed just a tool, and not a very sharp one at that. It's good for roughing out the shape of things, but I take that and fill in the details. I wouldn't ask it to make up anything other than expository text for descriptions, as in, "Take this paragraph and make it sound better." Of course, it has nonetheless generated useful new ideas that I've incorporated in, but that's been limited to some city names and details like notable religious groups. I don't use it if I can't fit it in with the other stuff that I've already written out myself to prevent such inconsistencies from appearing. To be sure, there are other generators that are not AI-based that are more useful in very narrow, tailored ways.

You're right in that it's not really reliable, which is why I am here with some of my stuff to get proper feedback from actual world-builders and designers. Ban put me up on a couple of new sites, but that donjon site is probably the best I've seen.

 

[Fyri]

Yeah! That is true!

If you do use Char.ai, I will warn you. You're probably gonna run into the same problems as ChatGPT eventually. I use it for roleplaying these days and the quality fluctuates.

 

[1MerryWriter]

They must have updated it because it's craptacular now. I ask it for expository descriptions based on (NOT plagiarizing) the text block I give as an example), and it just regurgitates exactly what I wrote. I've tested this now on new chats, and it's the same. It's like ChatGPT suffered a stroke sometime last week, and now, is nothing but an exercise in frustration.

This is to say nothing of the Sam Altman firing, near-rehiring, and sudden poaching by Microsoft.

Whatever they did, they borked it entirely. Now, I have to fight through twenty minutes of arguing with a dumbass (because that's what it has become overnight) to try to wrestle out usable information. Out of EIGHT paragraphs of plagiarized and mixed up bullshit, I got a three sentence paragraph. Wtf.

Are there any other ChatGPT sites that work just as good as it used to? Gotta be an alternative.

There is something interesting that I am noticing when I try to throw around ideas with CGPT. The paid version (CGPT4 - currently) does far better than the free version. The 'Creative Writing Coach' and some other CGPTs that are custom made to create worlds or characters, do much better with the more wonky ideas that you throw at it. It is likely a method to make people subscribe to the paid service more than anything else.
I haven't tried out other AI platforms specifically for bouncing ideas. The early attempts I made with Bard / Copilot haven't been that great. Most AI platforms, free versions at least just better for programming or for basic question answer type thing when you want a summary / overview of something quickly.

 

[Finchbearer]

It is likely a method to make people subscribe to the paid service more than anything else.

Yes.

Thing is, I’m always dubious about these editing platforms. I tried the free version of grammarly the other day just to see what its features are, but it pretty much makes it so it’s unusable unless you pay for a ridiculously expensive subscription. And then what are you left with? Basic grammar mistakes that as a writer you’d better learn by yourself rather than relying on tech to tell you if you’re wrong or not. My writing is far from perfect, but I’m not entirely sure using an editing tool would make it ‘better’.

 

[pmmg]

Funny. I did my own writing and the quality is better than both versions.

 

[Ned Marcus]

I've experimented a lot with AI for text and image generation, and played a little with video generation. Some of the images are beautiful, and it's good for brainstorming ideas for nonfiction articles but not writing them. For fiction, it's not even that good. Perhaps in ten years it'll be writing good novels—or perhaps not—but it's definitely not there yet.

 

[1MerryWriter]

I've experimented a lot with AI for text and image generation, and played a little with video generation. Some of the images are beautiful, and it's good for brainstorming ideas for nonfiction articles but not writing them. For fiction, it's not even that good. Perhaps in ten years it'll be writing good novels—or perhaps not—but it's definitely not there yet.

For fiction writing, it won't take ten years, considering the pace at which people are using the GPTs and feeding it information. The OpenAI version at least states that they don't use the data that people present in chat for training. Bard and Co-Pilot do use the data given by people when using their free versions. I haven't been able to test Grok so I don't know.

In case of non-fiction, the AI is just better than most writers. There was this whole fiasco with the 'Google Helpful content update' and its attempts to push back against tons of 'content' put out on the internet as a result of people typing in a few prompts and creating articles in minutes. AI generated content started ranking lower in google results but it hit quite a few people who wrote everything by themselves. Unless you have a clear expertise and a unique voice, you are not going to beat AI in non-fiction. You have to be creative and tell a story better than other non-fiction writers.

The custom GPTs for fiction do write 500-1000 word short stories better than most amateur writers. There is no accounting for taste of course and being objective when assessing the quality of creatively written content is difficult, after a point. I clearly enjoy what was written and worked on by another person but through a fair few discussions I have realized that most writers fall into this dubious category where one can't tell if something was human "generated" content or AI.

 

[pmmg]

For fiction writing, it won't take ten years, considering the pace at which people are using the GPTs and feeding it information. The OpenAI version at least states that they don't use the data that people present in chat for training. Bard and Co-Pilot do use the data given by people when using their free versions. I haven't been able to test Grok so I don't know.

In case of non-fiction, the AI is just better than most writers. There was this whole fiasco with the 'Google Helpful content update' and its attempts to push back against tons of 'content' put out on the internet as a result of people typing in a few prompts and creating articles in minutes. AI generated content started ranking lower in google results but it hit quite a few people who wrote everything by themselves. Unless you have a clear expertise and a unique voice, you are not going to beat AI in non-fiction. You have to be creative and tell a story better than other non-fiction writers.

The custom GPTs for fiction does write 500-1000 word short stories better than most amateur writers. There is no accounting for taste of course and being objective when assessing the quality of creatively written content is difficult, after a point. I clearly enjoy what was written and worked on by another person but through a fair few discussions I have realized that most writers fall into this dubious category where one can't tell if something was human "generated" content or AI.

So what do you want to be? A writer? Or a middleman to ai?

 

[1MerryWriter]

So what do you want to be? A writer? Or a middleman to ai?

I am already a writer. Just not that good

 

[pmmg]

Good answer. Least you mattered.

 

[skip.knox]

>In case of non-fiction, the AI is just better than most writers
This isn't really saying much, once you take Sturgeon's Law into consideration, so let's just take as given that most writing is bad writing. But the statement quoted is worth extending.

There are many types of non-fiction. If we're talking about something as simple as a weather report, there's hardly enough room to judge quality. Sports reporting might seem straightforward, but there used to be a real appreciation for outstanding reporting (journalistic) in that area. And of course, when it comes to crime or world events, well they award Pulitzer Prizes for that sort of thing. Somewhere along that spectrum, humans start to pull ahead of AI. The skilled ones, anyway.

In a different direction lies whole other areas of non-fiction writing, including biography, autobiography, scientific literature, history, and so on. With the entire gamut of scholarly writing, it should be evident that AI cannot compete simply because original research is a foundational component of that literature.

What we do with more hybrid forms--docu-drama sorts of writing--is very much up in the air. Well, so is most of it, really. I think it's much too soon to make reliable predictions or generalizations. Not that that will stop us from making them!

 

[Ned Marcus]

For fiction writing, it won't take ten years, considering the pace at which people are using the GPTs and feeding it information. The OpenAI version at least states that they don't use the data that people present in chat for training. Bard and Co-Pilot do use the data given by people when using their free versions. I haven't been able to test Grok so I don't know.

Know one knows this. It may take more to really copy human stories, or it might not happen. There's no way to tell.

In case of non-fiction, the AI is just better than most writers.

This simply isn't true, unless you're taking a vary low view of nonfiction writing. AI-generated content is generic and cliched, at best. I'm not just saying this; I've read a lot of it. It's very basic.

 

[Demesnedenoir]

You're right, because we are so far from actual "intelligence," it's crazy how people speak of AI. AI has been used in Photoshop for years, it just wasn't as advanced nor called that. Low-end writing is more vulnerable, but I'm not sure how one would define that.

Nobody really knows where it's headed. There's a wide range of possibilities.

Know one knows this. It may take more to really copy human stories, or it might not happen. There's no way to tell.



This simply isn't true, unless you're taking a vary low view of nonfiction writing. AI-generated content is generic and cliched, at best. I'm not just saying this; I've read a lot of it. It's very basic.

 

[pmmg]

This simply isn't true, unless you're taking a vary low view of nonfiction writing. AI-generated content is generic and cliched, at best. I'm not just saying this; I've read a lot of it. It's very basic.

This is probably true when you take into account the whole of humanity. By the time you get to a website like this you are probably in the top 20% already.

 

[Finchbearer]

Non-fiction isn’t academia, like just straight facts collated together - it’s opinions and an individual voice. If anything it would be harder for AI to replicate that, unless the prompt it ‘do it in the style of…’ and then obviously it’s just copying. Well it’s all copying.

 

[1MerryWriter]

Pardon the very long response. It is a lot of thought and it is in response to quite a few comments. I do go off on tangents but I have tried to edit it. (This is part 1/3) - Apparently I posted too much so the third one is awaiting moderator approval.

>In case of non-fiction, the AI is just better than most writers
This isn't really saying much, once you take Sturgeon's Law into consideration, so let's just take as given that most writing is bad writing. But the statement quoted is worth extending.

There are many types of non-fiction. If we're talking about something as simple as a weather report, there's hardly enough room to judge quality. Sports reporting might seem straightforward, but there used to be a real appreciation for outstanding reporting (journalistic) in that area. And of course, when it comes to crime or world events, well they award Pulitzer Prizes for that sort of thing. Somewhere along that spectrum, humans start to pull ahead of AI. The skilled ones, anyway.

In a different direction lies whole other areas of non-fiction writing, including biography, autobiography, scientific literature, history, and so on. With the entire gamut of scholarly writing, it should be evident that AI cannot compete simply because original research is a foundational component of that literature.

What we do with more hybrid forms--docu-drama sorts of writing--is very much up in the air. Well, so is most of it, really. I think it's much too soon to make reliable predictions or generalizations. Not that that will stop us from making them!

I believe we agree on this. I was saying that "Unless you have a clear expertise and a unique voice, you are not going to beat AI in non-fiction."
There was a space for an average writer or even an above average writer to work in public and hone their craft. Especially in non-fiction. Now that space will be filled with others who will 'semi-consistently' pump out AI mixed stuff. The key word for me is 'semi-consistently'. Many are too far away from reaching any level of consistency so most non-fiction writers who stick with it are still going to be fine. The world of weather reports and basic blogging / article-writing has long been "AI-augmented". I am not talking about that. I am talking about people who write anything longer than 10,000 words (or thereabouts) that can be called as non-fiction.

Say I want to write about some topic that is now becoming a hot topic but was not previously. I am picking non-fiction titles from top news stories here "The history of the land called Palestine / Canaan / Israel". Thousands of books have already been written but ChatGPT can write number 10,001 and make it happen in a few weeks even if the "writer" in question has zero idea.
The writer still has to fact-check, edit and must have some hold of the English language but it is now feasible to take such a topic and go from draft manuscript to self-published in at most two months. It would take two months if you wanted to do a decent job but if the writer is not even bothered with that, it is not going to take as long. Sure the returns might be low, but then the costs are next to zero here if one assumes that conventional writing time is something that any writer has to put in to a book.

If this person is someone with a social media presence and can do a bit of marketing - then the endeavor is already a net gain. I am saying this based on the academic ghostwriting market that I am very familiar with. There were quite a few educators of all kinds that got in on the early self-pub gold rush back when Lulu was still a name that people thought of. There was this 'startup' that I eventually chose to not work with which set up it operations at that time (2014-15 ish). Now they have hit the motherlode in terms of productivity and output.

Naturally, this does not concern creative fiction yet (as much at least).

AI-generated content is generic and cliched, at best. I'm not just saying this; I've read a lot of it. It's very basic.

True and as unfortunate as this sounds, there is a market for cliched non-fiction. Then there are also the people who know how to turn AI output into human output. If it alright to link medium posts (by others - not mine) I will link an article. Otherwise, it can be searched for on Medium. It was written in the end of December and in a publication that lets writers 'Practice in public' (no affiliations here). It is also called as such. This piece is suggesting ways for writers to use AI without making their content sound like it is AI generated. Naturally it has a lot of positive responses. (End of 1/3)