Bonjour,
Nouvelle règle sur le serveur :
Tout contenu généré par une IA générative (image, vidéo, audio ou texte) doit être clairement identifié comme tel, dès lors que vous en avez connaissance.
Bonne journée à toutes et à tous.
Discussion
Loading...
@contact en voilà une bonne idée !
@socialwg / @jonny @manisha @jdp23 and the rest of our instance at Neurospace.Live
Their server 👆 has a new rule:
"all AI-generated content (image, video, audio or text) should be clearly identified as such, as soon as you become aware of it"
Should we have something like this or even something harsher like "do not post AI-gen content in any form"?
@elduvelle @socialwg @manisha @jdp23 @contact i'm into it. Much like how we have exceptions for objectionable content for the sake of criticism, we should probably not have a blanket ban on all generative AI content in any form. Would be into drafting rules around this tho - i like the rule above as a starting point with disclosure, and would be down for extending it, it's getting late for me tonight, but maybe we start by gaming out the kinds of posts that wouldn't be welcome?
@jonny @socialwg @manisha @jdp23
Yes, that seems good!
I can think of the following situations:
- If a post is passing genAI as real, it should be reported and deleted (so we would need to as a "genAI" category for our reports)
- If a post is using genAI content with clear labeling and clear intent of it being the subject of discourse (e.g. mock, criticise, praise [if you really have to]), instead of using the content as source of information, then that would maybe be acceptable but the post should have a Content Warning.
@elduvelle @jonny @socialwg @manisha great suggestion, I'm in favor of it as well. We could potentially ask folks like Timnit, Emily Bender, Alex Hanna, and Ali Alkhatib if there are any good examples or best practices to use as starting places.
Thinking about it a bit more , one question here is whether we can realistically enforce this rule on the federated timeline. In principle it's possible: people could report when they see it, and we could hide individual posts and then limit remote accounts that post genAI more than once or twice.
But, it would potentially be a huge amount of work (including in some cases doing the legwork to figure out whether it really is AI-generated). So, hard to know.
@jdp23 @jonny @socialwg @manisha
True, I think applying these rules on our local TL would be a start and pretty low-effort.
To apply similar rules for all posts, we might need to restrict the rules to the clearly misleading posts. It already happens quite naturally that if someone posts a genAI photo or video without saying that it is genAI, some of the comments will figure this out, often with some kind of proof (in some cases it is hard to prove, but more often than not it is still pretty clear). So, relying on the community, I don't think it would be too difficult to moderate those..
1+ more replies (not shown)