PieFed Meta

1203 readers

13 users here now

Discuss PieFed project direction, provide feedback, ask questions, suggest improvements, and engage in conversations related to the platform organization, policies, features, and community dynamics.

Wiki

founded 2 years ago

MODERATORS

rimu@piefed.social

PieFed feature idea: AI content filter (piefed.social)

submitted 1 week ago* (last edited 1 week ago) by rimu@piefed.social to c/piefed_meta@piefed.social

26 comments fedilink hide all child comments

A way for people to screen out AI generated content, similar how people can screen out nsfw content already.

It works the same as the nsfw filter except underneath it's an integer instead of a Boolean.

Content authors will flag their content as AI-generated, on a sliding scale from 0 (no AI) to 100 (completely AI generated. Intermediate values could be used too although this amount of nuance may be confusing for some.

Users will be able to set a "AI generated threshold" which filters out content above that threshold. At the UI level this could be presented as a checkbox but maybe a slider would be good.

Mods need to be able to set the AI level on content, as they do now with NSFW content.

Communities will have an AI-gen value too, which is automatically applied to all content within. Instance admins can override this value too, for local or remote communities.

Thoughts?

you are viewing a single comment's thread
view the rest of the comments

[–] karasu_sue@pf.korako.me 12 points 1 week ago (2 children)

I use AI for all of my English posts, because I can't write in English myself.
I write in Japanese and have AI translate it into English before posting—this post included.

In cases like this, would I be expected to mark my post as AI-generated?
Also, is using translation to post generally frowned upon?

Maybe I'm just overthinking it though.

[–] rimu@piefed.social 1 points 1 week ago (1 children)

This would primarily be used for images.

IMO translation with AI is not worth putting a special filter on.

Also if someone with dyslexia or some other neuro-spicy condition needs a little help with grammar, that's fine.

There is a spectrum of AI use, which is why I proposed using a number for the value rather than a binary "Is AI" is "is not AI".

At one end of the spectrum could be an image made from a text prompt which would be considered 100% AI.

A photograph edited in Photoshop with the AI tool that replaces the background with a sunset could be flagged as 50% AI.

Maybe something I write and then ask ChatGPT to fix up the grammar and weird personal idiosyncrasies could be 20% AI.

Something like that.

[–] karasu_sue@pf.korako.me 1 points 1 week ago

I see—personally, I think it's a good idea when it comes to images.
While it's true that not all users will flag their content properly, moderators can still add the flag afterward.

It's one of those things where it's better to have it than not.

[–] GenosseFlosse@feddit.org 1 points 1 week ago

I would stay away from ChatGpt for translations, at least for English to German. In the latest iOS demo the live translation had "i need your help organizing my wedding" translated into German "i need you for my wedding". I have seen other examples with Chinese scam shops Auto translated which always stick out. Same with some games where the texts sometimes don't make sense.