Um. Not sure what else I can say. I’m not going to link to examples but can there be some moderation?
Yep, I generally agree, but manual moderation is unfortunately not feasible at the current scale. I have plans to improve the chat situation using AI (to detect trolls), but it is not a super easy task given that AI stuff is expensive, and it has to be economically efficient (since the comments plugin does not and will never require ads to power it). Banning VPNs from chat has helped a bit I think, because it’s harder for trolls to just switch IP address when they get blocked. For image generation, I have put a lot of work into it to patch loopholes for people generating illegal things, but it’s a constant cat and mouse game, so it’ll never be 100% unless I were to completely ban NSFW, which I will not do. Still, the situation here will improve over time.
Then there’s the problem of NSFW stuff coming up on the /generators page. For that, I have implemented some automated stuff to try to prevent it, but it’s not easy, because e.g. terms like “nude” can be used in anti-prompt of image generation, so it’s easy to accidentally flag SFW as NSFW. I also try to do a bit of manual moderation where I can, but it’s not really scalable. I will try to improve the situation here, and it will certainly become a lot easier to automate as AI detection stuff becomes cheaper and more robust, but it’s not something that I’ll be able to fix properly for probably at least a few of months.
Thanks for your help.
The thing that made me decide to say something was an image generator that had a style that had a combination of prompts that were really not OK. But I can’t see it on the list of recently changed generators now, so maybe you found it too?
I have some suggestions:
- Maybe you could make it so that people who put generators together have to say if there generator is SFW or not?
- Retire generators that are no longer managed (like if the generator owner doesn’t log in to moderate their gallery/chat)?
- Could there be a “Report Generator” button maybe to make it easier for you to find the ones that you need to deal with?
We are having a discussion on how we can solve this very problem in our genratiors so it doesn’t bleed over into other genratiors https://lemmy.world/c/casual_perchance for discussion about prompts picture shares and the like :) please join us
The devs are doing the best they can do sometimes stuff slips thru the cracks and people are exploiting those cracks the me and a few other users are trying to help fill the crack by giving the ai context about what to ignore in a nsfw prompt if it works we will try to submit it as a plugin :) I just wish they would understand the law that perchance is hosted and go elsewhere whare its legal
This has been a serious problem with Perchance ever since long ago. I remember when I saw a very bad word ON the first comments section in the
comments-pluginplugin page when I just started joining the community back in 2022, when the admin control, banning users, and the colorful tags feature hasn’t even made its way to the plugin yet. Probably has little moderation there, even to this day. But glad you’re reporting this issue, anyways!Also @[email protected], I think there should be some features for the default
generatorspage that could help prevent suspicious generators from being seen in the hub page even more, mainly:- Generator tags (red is suspicious/limited, yellow is mildly suspicious, green is safe/PG, and gray is unrated). I know there’s already a “banned words” for generator names, but that’s not enough. Red-tagged generators should have the thumbnails blurred or even hidden completely.
- Blocklist (block generators from users’ view)
- User-defined filter words (same as blocklist, but with user-defined keyword filters)
I can only think of these features, but most of these are already on VioneT’s Generator with Blocklist page, which is really an improvement. As for the gallery, I think that also needs an improvement in moderation and safety, too.
Also, for some reason, I usually find it kind of firey to discuss something like this. 😅
Sorry, I know this could be controversial. 🙇♀️
It’s not so much bad words that I’m worried about (bad terms maybe?). Swears are swears. But, there’s some things I think AI should not be used for (not going into detail, sorry). I dunno. I don’t want to be a part of a community that attracts that crowd for lots of reasons.
I will rephrase for OP:
I WENT ON GENERATOR X AND THEY HAD DONE LITERALLY AS MUCH AS THEY COULD TO BYPASS THE ANTIPEDO STUFF AS POSSIBLE SO IM SURE THEY ARE BRINGIN IN THE PEDOS OF THE INTERNET AND THEN THE PEDOS ARE INFESTING THE REST OF THE AREA AFTER THEY ARRIVE and to a lesser extend the gore, dom stuff, etc. WHYYYYYYYYY?
I DONT WANNA BE INTERACTING WITH PEOPLE WHO I KNOW ARE IN TO THAT AND THAT’S WHY THEY’RE HERE. IT WOULDNT FEEL SAFE FOR MY KIDS NOR ME.
deleted by creator
It’s hard to know what you’re talking about or provide any meaningful contributions if you’re too afraid to even name what you are talking about.
No worries. I’m here not only to add things up, but also to inform the Dev and other community members about this serious problem. That’s just what brings in the nature of AI in general, and you can actually give feedback/rating to help improve the AI Text Plugin in the AI Chat generator page as written here:
Your ratings help improve Perchance’s AI plugin, which powers this chat. Please do not submit ratings if your chat includes personal info.
And there are these buttons that are used to give ratings to the plugin (or the model that powers the plugin, at least).
You can also take a look at the Dev’s comment on not only privacy issues but also other interesting notes on the AI plugins, might be worth to take a look at. But sure, the moderation features definitely need to be improved so much more, especially on the AI plugins and the comments.
I could just read the title to kind of roughly know exactly the problem, but a description/detail to the issue would be greatly appreciated. It helps people understand more of your problem as they read more clues on what is happening and everything. 🙂
@perchance if you can give us a nsfw and a sfw tag then I’m sure we can make a antiprompt made for nsfw that would reduce the chances of illagel content being generated since you need to be looking for it to get it me and allo are exploring on ways we can help
