I remember trying it for some very basic TTRPG campaign prep and it gave me the most hokey by-the-numbers boring derivative shit I've ever seen. You have to keep in mind TTRPG writing is a genre of writing that mostly consists of ripping off books and movies you liked, anyway. And it couldn't even reach that very low bar. Genuinely concerned idea scroungers never heard of hitting the "random" button on TVTropes.
I remember trying it for some very basic TTRPG campaign prep
When I was GMing I really liked GPT-2 for just churning out some nonsense to fill in unimportant details on the fly while just riffing on ideas with my players to build sessions. Like sometimes I'd have a good idea for a run, and other times I'd just ask the players what sort of run they want and workshop ideas with them till we got an idea we liked, then I'd (openly) get some stilted and bizarre blurbs from GPT-2 to give a little backstory and flavor to that.
But that was also relying on how flawed and weird GPT-2 was and how well the absurdity of its gibbering meshed with the tone we were setting. I feel like if one were to try to use chatGPT for the same thing it would just be dull instead of producing entertainingly absurd nonsense.
It was useful to generate names for me, and also go more in depth about geography.
For the first I did things like: give me a list of fantasy names vaguely inspired by Ancient Greek/proto-IE/Sanskrit etc. Also based on characteristics, like “give me a name based in PIE that refers to black beard/holy land etc.”.
For the second I did things like explain vaguely the geographical characteristics for an area, then ask if that is realistic and how the climate and biosphere would look like taking x, y or z into account.
It worked… ok-ish.
For the first it made up a lot of fake shit. I even went on some rabbit holes asking for sources and finding some, in like French from 50 years ago, or not finding it at all.
For the second, I honestly don’t know. It was convincing? I don’t know enough about geography to really know tho.
I remember trying it for some very basic TTRPG campaign prep and it gave me the most hokey by-the-numbers boring derivative shit I've ever seen. You have to keep in mind TTRPG writing is a genre of writing that mostly consists of ripping off books and movies you liked, anyway. And it couldn't even reach that very low bar. Genuinely concerned idea scroungers never heard of hitting the "random" button on TVTropes.
When I was GMing I really liked GPT-2 for just churning out some nonsense to fill in unimportant details on the fly while just riffing on ideas with my players to build sessions. Like sometimes I'd have a good idea for a run, and other times I'd just ask the players what sort of run they want and workshop ideas with them till we got an idea we liked, then I'd (openly) get some stilted and bizarre blurbs from GPT-2 to give a little backstory and flavor to that.
But that was also relying on how flawed and weird GPT-2 was and how well the absurdity of its gibbering meshed with the tone we were setting. I feel like if one were to try to use chatGPT for the same thing it would just be dull instead of producing entertainingly absurd nonsense.
It was useful to generate names for me, and also go more in depth about geography.
For the first I did things like: give me a list of fantasy names vaguely inspired by Ancient Greek/proto-IE/Sanskrit etc. Also based on characteristics, like “give me a name based in PIE that refers to black beard/holy land etc.”.
For the second I did things like explain vaguely the geographical characteristics for an area, then ask if that is realistic and how the climate and biosphere would look like taking x, y or z into account.
It worked… ok-ish.
For the first it made up a lot of fake shit. I even went on some rabbit holes asking for sources and finding some, in like French from 50 years ago, or not finding it at all.
For the second, I honestly don’t know. It was convincing? I don’t know enough about geography to really know tho.