• If you were supposed to get an email from the forum but didn't (e.g. to verify your account for registration), email Wes at [email protected] or talk to me on Discord for help. Sometimes the server hits our limit of emails we can send per hour.
  • Get in our Discord chat! Discord.gg/stararmy
  • 📅 May and June 2024 are YE 46.4 in the RP.

Should We Allow Use of the AI in PMs?

Wes

Founder & Admin
Staff Member
🌸 FM of Yamatai
🎖️ Game Master
🎨 Media Gallery
As you may have seen Charaa's been using the forum ChatGPT Reply feature to self-reply in a thread and it's a little weird to see this as many of the posts and replies are getting deleted. Maybe it would be best if I changed permissions to allow for use of the AI bot in PMs to allow users to use this feature without making a bunch of reply notifications - the only problem is that while on the forum I can see what people are doing with the bot, but in PMs I don't have any way to make sure they're not abusing it. If you've used OpenAI's ChatGPT you know they're strict about adult content and certain other topics and I don't want anyone making them mad on my API account (that I pay for). What do you guys think?
 
Solution
I second that active staff get donator privileges. They donate a lot of their time and energy by doing what they do that I think they deserve it and it helps them be the best staff they can be imo.

I know at least one site supporter that has chat gpt reply to his rp and it's always like, "this is excellent high quality writing that I really like for x, y, and z" so it's money well spent from my secondhand experience!
Maybe you could make the AI Testing forum visible to staff like Charaa for their use instead of just site supporters. That would contain the AI stuff to a single place.

Cuts down on the visible replies for most users and there's no need for deletions because it's generally hidden and not bumping posts down. Though that probably won't help you see fewer reply notifications.

PMs would be fine and not even require a question but, because of the content considerations associated with your account, it's entirely dependent on your comfort level.
 
I admit I personally, delete in the hmm thread since I have them saved to a Word document, and dont want to use the prompts anymore, (until I realize I want to again) I also delete both in hmm, and in my rp posts, on discord cause I am worried of flooding. And in my business of the Mining Guild thread, since I need to do another reply so it added more. I have deleted it cause I wanted to merge the posts.

I’m sorry for any trouble I’ve caused!

As for the question, I’m going to have to say no, mostly because personally I don’t know the limits DMs allow, and ironically I don’t really want bot messages to be sent to DMs both on the forums, and on Discord.

Also, what kind of adult material is allowed?
 
Last edited:
GPT is very sensitive, especially the older model that is used by the bot. It self censors a lot of things.
 
I second that active staff get donator privileges. They donate a lot of their time and energy by doing what they do that I think they deserve it and it helps them be the best staff they can be imo.

I know at least one site supporter that has chat gpt reply to his rp and it's always like, "this is excellent high quality writing that I really like for x, y, and z" so it's money well spent from my secondhand experience!
 
Solution
"they're strict about adult content"

I mean that would have been my only concern to how it might be used but if its gpt/openAI stuff we wont be worried about ERP the dms with... Oneself?... I don't think i'll hjave a use for it, personally, as I barely use GPT-3 let alone having ever used 4 to even know the difference but for people like charaa and their thread it might be a nice alternative especially if someone wants to plan something or pitch something and doesnt want it seen on the forum first by everyone for whatever reason first.
 
I have to ask, since it may push me to use or not use. Is nudity allowed for the GPT-3? That's the extent for adult content i'm wiling to use it for.
 
I mean I use gpt-3 (I think... Whatever the basic gpt is.) and have used it to describe my most recent character as having a modest chest and generous hips or something. And i've never gotten a warning or anything tho I explain at the start that its for a scifi-based writing/RP community and I wanted it to help me with designing a character so unless you are getting into some very-very specific kinds of stuff I don't think you would have a problem and even then just following up with a reminder of the context usually suffices.
 
i'm assuming no sex, which is fine.

Its also weird about violence. I've had it occasionally advocate for torture and genocide and get pretty close to describing scooping out someones knee caps. But when I want it to describe what happens when I roll a 10 on the Dark Heresy explosive crit table "This content may be against our terms of service".

I'm not sure, it's not explicitly stated that I know of.

Its based off tokens, not character count. The limit is 8192 tokens in a single prompt. Its pretty huge last time I played with it. There is also a 32K token variant. I've not messed with my GPT account in a while.
 
Its also weird about violence. I've had it occasionally advocate for torture and genocide and get pretty close to describing scooping out someones knee caps. But when I want it to describe what happens when I roll a 10 on the Dark Heresy explosive crit table "This content may be against our terms of service".

Its based off tokens, not character count. The limit is 8192 tokens in a single prompt. Its pretty huge last time I played with it. There is also a 32K token variant. I've not messed with my GPT account in a while.

oohh, i dont know if its the same thing, but cool, if it is, would be helpful for stories. Tokens i mean
 
A token is basically a word. And it seems to actually choke on about half that much, and disregard everything after about 4k tokens
 
oohh, i dont know if its the same thing, but cool, if it is, would be helpful for stories. Tokens i mean
A token is basically a word. And it seems to actually choke on about half that much, and disregard everything after about 4k tokens
More or less what Yuki said, but more specifically, I'll use the sentence I just wrote and give an example what it might look like.

More/ /or/ /less/ /what/ /Yuki/ /said/,/ /but/ /more/ /spec/if/ically/, /I/'/ll/ /use/ /the/ /sent/ence/ /I/ /just/ /wrote/ /and/ /give/ /a/n/ /exam/ple/ /what/ /it/ /might/ /look/ /like/./

You can see common words are their own token, but longer ones are cut up.

And as Yuki said, its not great after a certain point. It parses each token 1 by one and updates a value thats about 400 bytes long before moving to the next one. This is called Byte Pair Encoding and can kind of be thought as GPT's memory, though if you play with it long enough you will find the limitation of 400 bytes pretty quick.

Though having originally been exposed to LLMs via the schizophrenic dreams of AI dungeon 2, it was a massive step up.
 
RPG-D RPGfix
Back
Top