What if labels could help with AI too?

[bovine3dom@masto.ai]

@dunesec also, the grades on the energ are totally nonsensical and not the kind of mistake a human would make

AъBCEFC

@digitaleu

[unkx@icosahedron.website]

@digitaleu sorry to be ”that guy” but could you tell us why did you delete the original post and all your replies there?

[cid_terron@social.linux.pizza]

@digitaleu shouldn't AI generated content also include the other labels??
Energy and Eco labels would make quite a lot of sense ...

[digitaleu@ec.social-network.europa.eu]

Hi @unkx thank you for asking. There was a typo in the original post, and when redrafting it we ended up republishing rather than editing, which removed the previous replies. This was a genuine mistake on our part, and we apologise for the confusion.

[digitaleu@ec.social-network.europa.eu]

A note from our team: earlier today we redrafted this post to correct a typo and refine the messaging, and ended up republishing it rather than editing, which removed the original comments. This was a genuine mistake on our part and we apologise for the confusion. Every comment and opinion is welcome on our page, and under this post. We value the conversations with our community.

[unkx@icosahedron.website]

@digitaleu fair enough. Mistakes happen and there’s no shame in them.

[joneskoo@infosec.exchange]

@digitaleu I like the idea of disclosure but not implementation (icons reference). Many categories are not going to be understood by content producers or consumers. Industry has already standardized on ”AI sparkles”, so introducing a text label that doesn’t translate to all languages seems like a mistake. Another cookie regulation?

[catraxx@tech.lgbt]

@digitaleu How about the energy label and the AI label get combined so we can see how much energy was wasted on making something that very likely already existed to begin with?

[boltx@mastodon.social]

@digitaleu Of course the image in the post would be AI generated without disclosure

[ko@gotosocial.art]

@digitaleu
>post about the need for an "ai-generated label" (good)
>the image itself is ai slop

holy fucking shit you lot are incredibly incompetent and dragging EU's reputation even lower

[nelson@wetdry.world]

@ko @digitaleu bruhhhhh i didnt even notice at first lmfao

[ko@gotosocial.art]

@nelson @digitaleu how long until the glorious european regime blocks me on The Mastodon Network for an opposing opinion

[srazkvt@tech.lgbt]

@ko @digitaleu hey technically the image has the label on it so it works ?

[karcsesz@equestria.social]

@ko @digitaleu I mean, the label *is* on the picture

[karcsesz@equestria.social]

@ko @digitaleu There is no point, just wanted to be a little silly.

[karmalakas@mastodon.social]

@dunesec @digitaleu they deleted the other post. This one is different. Probably too much to handle the comments. I guess they didn't know what the Fediverse is all about...

[dunesec@mastodon.social]

@karmalakas @digitaleu I've had a lot of experience with implementing EU symbols and guidelines. I was triggered by the wrong use of those symbols communicated by the EU. I'll give the social team some slack, because it is not their job to know every symbol and guideline, but I do wish they didn't use AI. Especially while talking about correctly labeling AI content. To be fair, those guidelines don't apply to this post, but it's sends the wrong message to me.

[madsenandersc@social.vivaldi.net]

@catraxx @digitaleu

There is no way you can determine the energy usage like that. Hate AI all you want, but please don't make others have to defend why the don't hate it as much as you do. Seriously. Don't.

With that said, I would really like that there are some pretty clear consequences if someone publishes AI-generated content without labelling it as such, at least when it is used for commercial use.

[catraxx@tech.lgbt]

@madsenandersc @digitaleu Sure you can, what makes you think that you can't? All you need to know is what model is used and then it's easy as soon as these companies are forced to reveal how much damage they are doing. Already happened with plenty of other products.

And i honestly don't care if people hate it less than i do. There are plenty of people who defend combustion engines, that doesn't make me want to protect their fragile egos either.

[madsenandersc@social.vivaldi.net]

@catraxx @digitaleu

Of course you can't. To the best of my knowledge there is currently no standard for measuring energy usage for running LLM's in any way, shape or form, probably beacuse it all depends on where the data center is located, how much the hardware is loaded at the time of inference and how cooling is achieved.

Even training is not a fixed cost. Depending on who you ask, training can make up anywhere from 10% to 40% of the total energy consumption, and considering that the average energy consumption per query for training is getting lower with every query, it's impossible to calculate on a per-query basis.

See this report for a deeper explanation:

https://www.sciencedirect.com/science/article/pii/S1364032125008329

Scroll down to this part and what comes below it:

"2.4. Discrepancies in estimations and underlying reasons
Existing estimations on LLM energy consumption, including both training and inference stages, exhibits substantial discrepancies, even for the same model."