In PieFed v1.4 it will be possible to label posts as having AI generated content. Labelled posts will get a little badge near their title, similar to how nsfw and nsfl content is labelled. Mods and authors can change this value on a post in the same way they do flair and nsfw, using the little tags button in the bottom right or in ‘More options’ when editing their post.
Account settings for blocking:
Similar to how NSFW works, each user can control how those posts are listed on the home page, etc. The default is ‘Label as AI’, which just adds the badge. People averse to AI-generated content might want to change this to ‘Hide completely’.
Also similar to NSFW, entire communities can be assigned as ‘AI content’ and that will auto-tag every post inside as being AI generated. If the community mods have this value unset (e.g. Lemmy communities, which don’t have this functionality) then the instance admin can manually override the community’s AI Generated setting.
This is excellent! Will there be a feature for users to report a non-tagged AI post to mods for tagging?
Could you use the existing reporting feature for this?
Holy moley I love Piefed so much
And it just keeps getting better, every month!!:-)
Only improvement I would suggest is to make the label “GenAI” instead of just “AI”
Even more, I’d like to see some technically correct term, instead of the “AI” buzzword that the companies are forcing with their marketing
(it’s not intelligent, it’s just a token generator/predictor, similar to your phone’s autocomplete feature, just more advanced and trained with everything in the world)
Unfortunately I’m not an expert, but maybe something along the lines of “Deep Learning”, “Neural Networks”, “Generative Networks”, “Generative technology”, etc
You’re right, it’s not really an accurate term - but it’s the shorthand abbreviation that has caught on.
GenAI stands for Generative AI, which I think is descriptive enough
I think their complaint is more that it’s an umbrella term applied a bit loosely (like plain old “AI”). Most people, even non-technical people, seem to instinctively understand that there’s some sort of machine learning (or a novel type of less rigid processing) behind their chatbots and image generators, but most of them also seem to think it’s one artificial entity that does the whole thing. So non-technical ChatGPT users are likely to think all of the different features (like asking it in text to generate an image) are one “consciousness” instead of a text environment and an image environment that sort of interface with each other. And then each of these is made of different parts. It’s all one product (I think we both agree it shouldn’t be one)
I think the person you’re replying to doesn’t like the term for the same reasons I do, it’s quite blatantly just a marketing term, a buzzword to swat away like an angry wasp. But it is the most descriptive and succinct term that most people know. I’m quite conflicted on this. “Taking a few pulls at the slop machines” is a fun way to shit on someone else abdicating their responsibility by using machine generated content en lieu of their brains, mostly in a context where their brains are needed, but “slop machine” is too accusatory. In the same way, I find “GenAI” just a little too permissive or possibly enthusiastic about what is claimed these algorithms even do.
If it was totally down to me I would probably add multiple tags: “Machine generated text”, “Machine generated images”, “Machine generated audio”. It keeps the distinctions I think are necessary (text ≠ thought/analysis, image ≠ artistic expression, etc) and gives us a few more toggles and a bit more statistical information that would be nice.
This comment dragged on a bit longer than I’d set out to write.
I think the issue is that it’s not that much generative nor intelligent
SC - synthetic content
Names the problem without relying on their marketing
That seems vague though: when is something synthetic? If I use something like Context Free to create something, would that count? If you paint or take a picture, but apply some filters / effects? It, to me, feels too general to differentiate between tools that “fill in based on training data, and statistics thereof” and those that do not.
While AI/GenAI is used for marketing, it does draw a clear boundary as to what necessitates the annotation, yet is general enough to include any future changes, while being widely recognized.
If I use something like Context Free to create something, would that count? If you paint or take a picture, but apply some filters / effects?
No, because what we see is not a randomized output based on statistics of previous content
But 😅 my point is mainly about using a correct technical term, instead of the forced marketing buzzword “AI”
Another piefed win
It has been a big weekend.
There were 32 commits and only a third of those relate to things announced in this community. There’s some pretty nice integration with Mastodon, Pixelfed and Loops coming.
Omg I’m blind. I read that as “AI images will get a small badger next to the title” and I was so confused, why a badger, what do the cure buggers have to do with AI.
I mean… be the change that you want to see in the world: write that code that will insert cute little badgers next to (haha, random?!:-P) things all throughout the Threadiverse!
You could put it on a custom instance. Maybe call it “badger badger”?
Which mobile apps are keeping up with changes to Piefed?
Pretty much none, since a lot of the best features of PieFed are not (or at least historically were not) exposed in the API for them to tap into.
PieFed is moving forward so blazingly FAST that it might not seem like it, but overall it has not had the same amount of years of effort put into it that Lemmy has. Which imho makes it all the more impressive how its feature set has already (mostly) eclipsed that of Lemmy, and as we see here in the OP, also eclipsing that of Reddit as well. (But not in all ways, and the mobile API and the content searching are the two main ways where PieFed is a bit behind.)
That’s why I made an AI generated community, because a part of the solution to stop people from posting AI generated stuff in other communities is to give them a place to post it. It still has to follow all the rules, but worst case is that someone posts something real in the AI generated community. https://lemmy.world/c/AIGenerated
Good idea but it won’t help with accidental AI posting - when OP genuinely didn’t know that the image was AI generated. The label can be set AFTER posting which is good.
The unfortunate thing that needs to be recognized is that we’re already beyond the point where someone skilled can create AI generated content that looks hand made or like a real photograph. The thing we need to be focusing on is context and corroboration. Even if some content looks completely real and genuine, if it isn’t corroborated and no one knows where it came from, then it becomes safer to assume that it was generated with AI, manipulated, or photoshopped.
Well, we tried that with [email protected] , the success is limited
People will never follow the rules. I mean some will but you can never count on all doing so.
Inside of AskUsa most posts from non-mods are on explicitly political topics, despite efforts to state many other communities (stated by name and with links provided!) where those should go instead of there.
And then outside of that community people refuse to put their questions into it.
Part of the problem may be how the sidebar text tends to be fairly or even extremely hidden by some mobile apps. It would be helpful if the create post screens would show that - on all platforms: PieFed does, Lemmy does not, and some apps might or might not depending on the choices of those particular devs.
People will never follow the rules. I mean some will but you can never count on all doing so.
Indeed
This isn’t going to work out. A lot of people that post AI made content are of the kind that don’t know it’s AI, or simply don’t care. They’re not going to tag it.
Then mods will tag it.
The mods can label them
will they? mods who are okay with AI content wouldn’t care. mods who don’t like slop would just remove the post.
People can block an entire community, preferring others instead where the mods keep out those kinds of posts. Nothing is perfect but this is a new tool that has potential to help people get what they want. e.g., is there anything wrong with NSFL posts, or bots, so long as they are properly labeled as such? If this tool helps organize at least some of the mess, then it can be helpful some of the time?
Ofc, only in PieFed communities, and for people with PieFed accounts. Though if Lemmy likes the idea then perhaps it can add it as well (same for Mbin, nodeBB, flarum, pixelfed, Mastodon, Loops, etc.).
Awesome! Thanks for this.
Dark pattern design.
First of all, you’re calling chat bots / image generators AI, which is just free marketing for Nvidia and so-called “Open” so-called “AI”
Laymen don’t understand that Google Translate and ad algorithms have been called “AI but not AGI” for years in the same sense as these chat bots. The massive surge in use of the term “AI” is purely meant to confuse laymen and sell them the idea of intelligent bots we don’t actually have.
More importantly, by using a “this seems like a bot to us” blacklist instead of a “we have proof this is a human” whitelist, you’re fucking up democracy. If a system like this has enough users, it will probably end up being the direct trigger of suicides by people unfairly labeled as bots (like a non-proof-based whitelist would cause with people unfairly blocked from the list).
I hope this is a mistake, and not an intended step in the enshittification of the internet.
it will probably end up being the direct trigger of suicides by people unfairly labeled as bots
Yeah, possibly? Then again, the same could be said for running out of milk at the local grocery store, and more generally not everyone you dislike is (literally like) Hitler. By framing your argument in these terms you are distracting from the goal that you seem to be trying to reach. Take my answer however you will but I meant it in kindness.
I don’t see how a proof-based whitelist would systemically be the trigger of any suicides.
GenAI labels should always default to “human-made”, with labels applied to generated stuff, not human-made. Also, being wrongly accused of using AI by a mod isn’t much different from being banned over a misunderstanding. It just means we need multiple mods per community and a full transparency of mod actions.
Personally, I see more benefits from having genAI labels.
Being isolated because you’re not good at proving you’re human is different from being banned over a misunderstanding
You aren’t isolated though, it’s just your single post that’s been labelled as AI-gened, and by PieFed doesn’t even hide the post from users. Also, I disagree that it’s any different: being banned because people thought you were bigotted when you didn’t mean to can be a brutal experience.
And if one mod/admin keeps accusing you in using AI then call out their behavior, avoid their communities and move on. Sometimes people are just assholes.
I think we have different visions of how chaotic the internet is going to get with bots aggressively trying to take over. The issue would be actual social isolation due to bots and glowies successfully convincing people your human work isn’t real. You make music at a time when human musicians are more focused on sounding human than sounding good; your song is really good; everyone calls it AI; you end up homeless and drifting away from the small number of limited human connections you had. Others with more social connections (and an easier time proving they’re human) are able to peddle actual AI music as their own. Absolute dystopia
If it’s just the occasional mod being an asshole, that’s not so bad
I can’t see how labeling a post as AI-gened in PieFed is gonna make you homeless, sorry. Unless you’re talking about the AI impact in general but then I don’t see how it’s related to your criticism of this feature.
Tagging posts as AI-gened is good NOW, not in a hypothetical dystopian future where people abandon you because your music was accused of being AI-gened. Maybe that future will never actually come, for whatever reason!
My fear is it’s coming and this is put in place first to make use of it for dystopian purposes.
If we’re lucky, it’s not coming, or this is just a mistake that will be fixed before becoming a real problem.
No, it’s not put in place for that reason. It’s put in place to try and label AI content and allow users to filter it out.
I guess we have to disagree about whether this feature is dystopian or not. Have a nice day!
