Meta-commentary on this discussion: okay, I know the ‘hacker news’ thing has always been kinda a misnomer, and has gotten moreso. But it’s still kind of shocking to me how little people here value making something, versus pretending to make something. There’s value to actually doing stuff.
The amount of “all artists will use AI to pretend to make stuff, because money” is honestly kinda bizarre, and completely misunderstands the motivations of most artists, writers and makers.
How do you define "trying to make it look authentic"? You make a deepfake of a politician saying a discriminatory joke, and the author will just say "It was obvious this was not intended to be authentic. Anyone could see it's out of character". But the message already influenced millions.
The law is phrased “real people” which doesnt mean people that exist or are famous but rather people in general. So ai images of people using your product or whatever are real people which have to be labeled.
> ... who use an AI system to generate or manipulate image, audio or video content that appreciably resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful (deep fakes)
> ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments
To me, this definition reads as only including "agentic" uses, or automated decision engines. An image generator does not operate with any autonomy. If you interpret the "input" part loosely enough to include it, then you'd likely include old and simple techniques like seam carving [1], which very few would consider "AI" in the contemporary sense
If it looks like slop, it’s obviously AI. If it looks like someone actually put some effort into it, or if it looks like “real”, then it needs to be labeled because you wouldn’t know it was AI otherwise?
Most of the media you see in the near future (if not already) is going to be at least AI assisted. They're going to be overloaded by reports of AI generation not labeled as AI.
I'm also curious, what about AI assisted media? If a company writes a full article, then uses AI to edit it a bit, does that need to be labeled?
Smartphone cameras also do a bunch of image processing that can be classified as AI, by default (some more than others). I wish they'd stop doing that, but I'm sure they won't.
I agree it won't work, albeit for an entirely different reason. There has been a shift towards labelling AI generated or AI assisted content by those who aren't attempting to mislead. Those who are attempting to mislead won't.
As for AI assisted edits, the article is quite clear about the labelling covering AI generated content. If you "hire" the AI as an editor by asking it to check your grammar or point out weak arguments, then do the rewrite yourself, it almost certainly wouldn't be covered. It sounds like it wouldn't be covered if you asked it to apply edits for style either, just be sure to log the changes since it would clearly be in an AI's voice.
Sure, but this mandate means that if they're caught (which isn't unlikely), they can be fined up to 3% of revenue (not profit). Many will be outside the EU's jurisdiction, but not all. Taken together, this will have a chilling effect, not stopping it entirely, but likely reducing it.
It could turn out like the California “chemicals present that could cause cancer” that are ubiquitous in California, and therefore not useful as signal.
Maybe. But the rest of your post doesn't have anything to back this up.
> I'm also curious, what about AI assisted media? If a company writes a full article, then uses AI to edit it a bit, does that need to be labeled?
Why would it? The threat is mass generation of articles without any human oversights, and realistic looking images claiming occurrence of events that never hapened. Nobody cares about an AI typo and stylecheck.
Skip to the end. Let the consumer verify that the product is the unmodified intention of the entity claiming to have produced it and let them decide if they trust it.
Tangent: it is notable that certain variations of labelling of that sort would now be considered unacceptable by many:
"Let the customer verify if the flour has asbestos in it, and decide if they trust it."
"Let the customer verify if the car airbags were installed properly, and decide if they trust it.
"Let the customer verify if the plant that produced the plastic in their grocery bags is dumping its waste products in the water nearby, and decide if they trust it."
No.
We have governments and regulations for a reason. It is utterly unreasonable to expect customers to verify everything themselves. We need clear, firm labeling laws for all kinds of products. There's no reason this should be different.
You will have to tell me how "labelling machine additions" is applicable at all as an example here. Unless your argument boils down to "computer doing things does not require labelling".
That is mostly my argument. I find "computer doing things" generally not a useful category and why here specifically doesn't have enough support. If there are some truly high risk gaps, then why no close them specifically?
Hard to assess what this rule will actually accomplish. Ultimately AI will be involved in all creative processes going forward, so all pieces of media will have at least some basic AI disclaimer. At the same time, the bad actors that the EU is so paranoid about (Russians secretly undermining our democracy and other boogeymen) will obviously not use the label.
But it's another tool in the EU's toolkit to arbitrarily enforce fines on big tech when politically convenient, of course. That's one thing we remain good at.
With enough money you can already buy attention and presence. Now you can even facilitate a fake discourse to claim legitimacy and steer manipulation into entirely new realms. If this is not counteracted not only will rogue nations destroy democracies even faster, but powerful corporations and rich individuals will also use this to increase their power further and damage simple consumers.
More authentic because of what, a missing watermark? It still has 2 positive outcomes:
1: It gives legal leverage against vectors of misinformation. Misinformation is tricky to combat legally and this is a first obvious measure that can be taken.
2: Reputation will become more important in the future, and if you are found to have repeatedly posted AI content claiming it is not it will damage you. Not to mention that users can point out the obvious discrepancy.
> Ultimately AI will be involved in all creative processes going forward
What on earth makes you think that? Do you honestly think that literally all creatives will just start making lazy crap because… it’s cool? Because I can assure you, outside of this little corner of the internet, it very much is not.
Proper authors, journalists and artists are, generally, not keen.
There's creativity and there's the creative process. I like to dictate documentation in long form monologue, and then give instructions how to structure it. Then I give it a final pass with manual adjustments and the end-result is exactly how it would have looked if I had written it from scratch but without the bsywork.
Don't be stupid enough to let machines do your own thinking, treat them as a dishwasher.
So what are journalistic texts if not docs? At this point, an LLM is just another evolution from the typewriter.
You are still responsible for your work. What is important are the semantics and statements of the text. Less so how the exact arrangement of letters came to be.
"Even I - with all my years of experience and knowledge - fell into the trap of hallucinations. I summarised reports using AI tools and worked from those summaries, trusting they were accurate.
He used an LLM to perform research instead of using it as a mere writing tool. I think large language models are great at formatting text, formatting, structuring and giving a stylecheck. But the last pass and review must always be a human. Everything posted by an LLM under your name is still under your name and you are liable for every single word. It's about the outcome and everybody who uses it is responsible for it.
In this context by proper authors I mean authors published by real publishers, for a start. Generally, most are hostile to LLM writing.
Also, like… why do you think people write books or make art? In the vast, vast majority of cases it is not to make money (virtually none of them do). It’s to _make_ something. Where’s the feeling of accomplishment in making them if you just told an LLM to shit something out? Like, who would bother?
Publishing is a business. There will be real publishers who are interested in using AI for the sake of making money. Even when a publisher has integrity and refuses to sell AI generated books, they have to contend with authors who are only interested in making money.
I'm not sure if I'm using the right term here, but publishers have been using rotating ghost writers to write series for generations and many of the authors have been happy to translate outlines into books just to earn a living. (I know of ghost written series that my father read as a child. I'm by no means young, and he had children late in life before it became fashionable to wait to have children until they were older.)
> In the vast, vast majority of cases it is not to make money (virtually none of them do).
If you're suggesting that virtually no authors write books to make money, I would suggest that is not true and the landscape is going to change for the worse. There are constraints on how many authors write books purely to make money because writing takes time and effort, with the returns being notoriously low. AI eliminates the time and effort factor. Returns don't have to be high if it is being generated by a content mill. If anything, diluting the market will result in lower returns for human authors -- discouraging them from writing in the first place.
> they have to contend with authors who are only interested in making money.
This isn’t _really_ a thing. If you’re into writing for the money, you’re probably in the wrong business. Even some quite successful/popular authors have a day job; only a very few can support themselves on it.
My apologies for a lack of clarity in that sentence, but I meant that publishers have to contend with people who are willing to use generative AI to make money by writing books for them.
That said, real people write for all sorts of reasons. Money is definitely one of them. Sometimes it is out of ignorance (e.g. they don't know that few books earn much money). Sometimes it is out of delusion (e.g. some authors are clearly raking in the money, they think they will too).
> Where’s the feeling of accomplishment in making them if you just told an LLM to shit something out? Like, who would bother?
Exactly. Find something else to do.
From this perspective the mere existence of AI tools for this stuff devalues all competing possible results, whether the tools are used or not, simply because there will be so much possible everything (media wise) on demand to match the whims of any consumer so they lose interest.
Surely a better idea would be to focus on how we can idk, get camera firmware to cryptographically sign photos or somethin, and then have editing software apply a signature on top. Or at least encourage a system where theres a good chain of image provenance for photojournalists
People have already tried this with C2PA, and it's not a perfect system. The main problem with these systems is that you risk certifying fake media, which is worse than uncertified fake and real media.
I do wish more people used basic watermarks on their media regardless - as in sloppy ffmpeg text. People are more likely to repurpose other people's older media than generate new media, and old time-y watermarks make that just slightly harder. Maybe a free app for uploading citizen journalism media that handles the watermarking with geo- and datemarking is the way to go.
This is honestly not a bad idea. And could be done in parallel.
I even think that it might not need a mandate, if camera manufacturers start making it as a marketing claim. Prove your work is authentic and not AI generated!
This misses the forest for the trees. The important question is digital provenance so you can verify it came from the entity it claims; whether the entity you trust used AI is not the point. Plenty of people were not to be trusted before, only the EU would insist you trust and distrust who they say you should.
This is going to get very interesting, I could see some erroneously claiming something as AI and demanding that I have an AI label when it's not. I'm not sure how this is even enforceable.
- law doesn't play "wording nitpicking", "Simon said" or similar nonsense. And this applies here, too.
- law doesn't care about what you or any "specific" person thinks something means/looks.
i.e. if you ask 1000 random people if something "looks authentic" and more then a very very small fractions says it is "authentic looking" then it is authentic looking (idk. what "more then a very small fraction" is, but 10% is definitely more)
Honestly any honest interpretation of "do you try or thing a random person from the street might mistake it as authentic (based on the _looks_, not content)" is giving you a pretty good idea for what it means. (honest here == not definition nitpicking, not egocentric, not elitist, not defaultist, etc. )
But also most relevant:
_This (authentic looking) is the wording of the article not the law/regulation._
This is mostly about content which tries to manipulate people with deep fakes.
And stuff like it being clear if e.g. a AI backed voice translation was used, as there had been more then one or two cases of voice translations being wrong in a way you would attribute to malice for a human translator.
Another step to make users dumb and blindly trusting of their regulators. I don't opose labeling AI content, I opose mandating it and giving users the false confidence that something without label is legit.
Let's start labling everything. Let us label every cigarette individually, every single drinking glass, every single nail.
No, people still smoke, drink and fornicate, despite all the labels.
At some point, we should just let Brother Darwin sort out the thinking ones from the non-thinking ones, because at a too aggressive level of protection, society will stop working, and people will regress to children.
> Participants drew on a range of risk discourses in framing their abstention or moderate drinking along the lines of health, wellness, wisdom and productivity. They reiterated social constructions of heavy or regular alcohol use as irresponsible, threatening and potentially addictive.
The non thinking ones have voting power too and they will burn this shithole down because they think some powerful people in nice suits act on good faith and should be given all the decisionmaking and resources.
This way of thinking is a very efficient way of destroying your culture and civilisation.
If I buy a bottle of, say, white spirits, there’s a warning on it that it’s poison and dangerously flammable. Because that’s the law. Now, there’s nothing _really_ stopping an unscrupulous manufacturer from making a bottle without those warnings. But on balance it still feels preferable to force manufacturers to do it.
My biggest problem with this is that this regulation is only aimed at AI. It should be much broader and include any technology that can be used to make 'authentic-looking' content. Why should someone who's good at Photoshop or VFX be allowed to share fake stuff as real, but not someone who's good with AI?
For many years we've been aware that edited models on magazine covers, Instagram and in ads are negatively impacting the well-being of (young) people, but we've never done anything about that. I consider this regulation incredibly hypocritical and full on populist because 'AI BAD!' and nothing else.
In Norway ads already must specify if a person in it has been digitally altered (changed shape, smoothed skin etc). Quite funny with international ads getting these banners plastered over their face. Random example: https://imgur.com/a/t8zT37o (rules for size and contrast)
AI is capable of generating things that are nearly indiscernible from reality, with almost any input, be it adversarial or other. Photoshop at least involved some manual trimming and lengthy process that couldn't be applied to "literally anything you can thing of".
I'd rather have the laywork already there to punish sensationalist publishers that abuse AI to forge alternate realities than nothing.
Manipulating information is going to be the plague of the post-GPT/post-Midjourney era. We are entering an era where trust if going to be hard earned, and I don't know how you can trust for-profit or politician/billionaire owned companies anymore. What prevents WaPo from manufacturing information?
What about the practical difficulty of defining "AI"? You could explicitly call out, say, diffusion models, but that would leave future holes for new AI technologies
What about Photoshop being driven by Claude Cowork? That my seem asinine, but it is a serious consideration
I found this definition here [1]
> ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments;
But that barely clears anything up. Even worse, I would not include a simple user-prompted image generator under this definition, as it is not operating with any autonomy. Or, you interpret it so broadly that seam carving [2] is considered AI, let alone content-aware fill
"It is a matter not only of customer protection, it’s also a matter of democracy protection."
Ensuring democracy by selectively censoring. I feel safer already.
I better put AI label on every video I make as there is already AI in the supply chain of all the tech I use. All content contains AI now.
Let's talk about AI commissars who will decide which company will survive.
In fact, let's create a few an NGOs to help! NGO will tell commissar what he wants to hear, NGO will get rewarded and allowed to exist!
I doubt EU will survive the next decade or two without major crisis or war tbh.
Edit:
Instead of downvoting, how about you try retorting?
I don't think I'm incorrect here. But maybe I am!
I am seeing more and more this exact sentiment that is in this comment: "I should be able to bamboozled everyone into thinking I have skills I actually don't have anymore".
Look, there's a ton of places where you can display your AI code, products or comments. Why is it so damn important that no places remain for people who want to be aware of whether or not they are engaging with a human?
Why is it so important to you that users must not be aware?
My suspicion is that it will quickly not matter in a lot 9f situations for most users, customers, etc. - so might be easier and better to curate places of true human interaction, for example, instead of the other way around.
Slapping an explicit "I did not put effort into this" label on slop content is not censorship, its quite opposite, it lets people filter out the noise. No human can hope to be able to do this on their own at the scale GPUs can pump out tokens.
Authentic voices are today being drowned in waves upon waves of AI slop spam. This is not quite censorship but the end effect is similar - human voice cannot be heard. Just this time not because someone stopped the human from talking, but because they as shouting 1000x louder instead, at pennies on the dollar it costs to create something truly valuable.
Probably not critical, just helpful to be able to understand where the source is willing to cut corners. Helps me understand if the given content source values are aligned with mine without risking wasting time to ingest slop just to come to the same conclusion.
The same way I look at food labels to support local produce and those who don't try to trick me into thinking I am consuming something it's really not.
I personally wouldn't care for AI usage in icons, I don't put a lot of attention to them. Someone else might.
Buying a license for a stock photo vs using AI to create what you need is cutting corners? Don't thing there is much difference in effort (might even be more on the latter)?
Buying a license for a stock photo is paying someone who has put the effort.
Well, okay, I am not sure how much effort goes into creating your typical stock photo, probably not a lot, just still more than 0 effort it takes for the AI to generate it.
And yes, I see difference between an automated system flooding the web with AI generated "news" in disguise vs a niche blogger using AI to iterate on an image with hand crafted prompts. Labels are not perfect, just better than nothing.
requiring deep fakes to be labeled as deep fakes has nothing to do with censorship
there is also nothing selectively about it, it applies to all deep fake equally
if you had read the article you would know that this is mainly about deep fakes and other deceptive content, not every use of an AI tool (e.g. line smoothing) requires an AI label, specifically it's only about content "generative AI" usage _in a context where it's potentially deceptive_.
> AI commissars who will decide which company will survive
a huge jump in logic not related to anything in this article
> I doubt EU will survive the next decade or two without major crisis or war tbh.
again unrelated, but any country which doesn't fights deep fake propaganda will most likely not survive as a democratic state of law, soo if anything this action increases the chance of survival
> Edit: Instead of downvoting, how about you try retorting?
- then maybe read the article properly, your comments put it highly in question you did
- also maybe don't have non logically sustained jumps in reasoning in your article, especially not when you jumps in reasoning come with tons of sarcasm
> I don't think I'm incorrect here. But maybe I am!
well yes, you have even missed the long known, explicitly pointed out sever threat deep fake propaganda is for any state and especially democracies. That is not a good bases for any argument.
Also passive aggressive edits are a good way to get even more down votes if anything.
Content label doesn't transform the content. It just provides a false sense of category.
I still have to think critically whether there is a label or not.
Cons: New ways for sociopaths for corruption, embezzlement and censorship.
That is my point.
This idea that legislation has no downsides is extremely naive and borderline retarded.
Look at proposed legislation from the point of someone who wants to destroy everything. Does it provide a path to that goal? Is the risk worth it?
Every government grows over time and produces more and more suffering while consuming more and more wealth. I recommend reading The Gulag Archipelago. Your thinking seems to be dangerously naive to you and people around you.
I think there is no alternative that yields good results.
Legacy media uses the argument that consumer already expects to be lied to and knows the content is for entertainment (TV news for example). I feel the same about AI content.
Or should I be expected to blindly trust everything? I already have to critically think
and filter everything already so why the distinction? I should try harder for some content and not another? Especially as AI gets better the distinction between AI and no-AI generated content gets blurrier.
All I see is bureaucrats trying to grab more power over people's decisions and more money embezzled. Unelected people forcing rules on their subjects.
Nihilism is the path to death and destruction. Thinking this will makes you powerless and cedes control with those who have the most resources.
It's easy to take yourself out of this equation, but it only makes everything worse. The trick is that they make you think that it couldn't get worse. But this is always false.
Can you elaborate? I have hard time seeing nihilism here.
(there may be language barrier and/or definition issue at play - maybe?)
With definition of nihilism along the lines of `the rejection of all religious and moral principles, in the belief that life is meaningless`
The point I am making (or trying to) is quite the opposite. Bureaucrats seem to be introducing more violence and even more problems while resolving none. I still have to critically about the content I consume regardless of it having some label. But now tax money goes towards this instead of the things citizens voted for.
You think that the status quo sucks and that there is no alternative, that's your prerogative. But why do you complain about people who think there is an alternative and try to implement it?
Meta-commentary on this discussion: okay, I know the ‘hacker news’ thing has always been kinda a misnomer, and has gotten moreso. But it’s still kind of shocking to me how little people here value making something, versus pretending to make something. There’s value to actually doing stuff.
The amount of “all artists will use AI to pretend to make stuff, because money” is honestly kinda bizarre, and completely misunderstands the motivations of most artists, writers and makers.
The problem isn't the "real" artists, it's everyone else.
> There’s value to actually doing stuff.
Maybe sometimes. But rarely. Usually value is in the actual stuff.
How do you define "trying to make it look authentic"? You make a deepfake of a politician saying a discriminatory joke, and the author will just say "It was obvious this was not intended to be authentic. Anyone could see it's out of character". But the message already influenced millions.
The law is phrased “real people” which doesnt mean people that exist or are famous but rather people in general. So ai images of people using your product or whatever are real people which have to be labeled.
Look at the written law.
I found this definition here :
> ... who use an AI system to generate or manipulate image, audio or video content that appreciably resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful (deep fakes)
https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
That text defines AI as such
> ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments
To me, this definition reads as only including "agentic" uses, or automated decision engines. An image generator does not operate with any autonomy. If you interpret the "input" part loosely enough to include it, then you'd likely include old and simple techniques like seam carving [1], which very few would consider "AI" in the contemporary sense
[1] https://en.wikipedia.org/wiki/Seam_carving
"Varying levels of autonomy"
So from image generators up to agents.
By the effect it has, among other things. Courts take into account whether the stated motive is believable.
If anything this an argument _for_ clearly labelling manipulated photos and AI generated depiction of supposedly real life events.
Yes, but without exceptions for satire and parody etc.
I'd argue that if a satire or parody is strong enough having a small watermark will not detract from its statement.
If it looks like slop, it’s obviously AI. If it looks like someone actually put some effort into it, or if it looks like “real”, then it needs to be labeled because you wouldn’t know it was AI otherwise?
the law will allow normal broadcast outlets to filter; so some of that will just be the pipeline.
If you're talking about twitter, then yeah, that's a firehose of crap and there's no laws that'll stop the "owww, it was just a joke!"
It won't work.
Most of the media you see in the near future (if not already) is going to be at least AI assisted. They're going to be overloaded by reports of AI generation not labeled as AI.
I'm also curious, what about AI assisted media? If a company writes a full article, then uses AI to edit it a bit, does that need to be labeled?
Smartphone cameras also do a bunch of image processing that can be classified as AI, by default (some more than others). I wish they'd stop doing that, but I'm sure they won't.
I agree it won't work, albeit for an entirely different reason. There has been a shift towards labelling AI generated or AI assisted content by those who aren't attempting to mislead. Those who are attempting to mislead won't.
As for AI assisted edits, the article is quite clear about the labelling covering AI generated content. If you "hire" the AI as an editor by asking it to check your grammar or point out weak arguments, then do the rewrite yourself, it almost certainly wouldn't be covered. It sounds like it wouldn't be covered if you asked it to apply edits for style either, just be sure to log the changes since it would clearly be in an AI's voice.
> Those who are attempting to mislead won't.
Sure, but this mandate means that if they're caught (which isn't unlikely), they can be fined up to 3% of revenue (not profit). Many will be outside the EU's jurisdiction, but not all. Taken together, this will have a chilling effect, not stopping it entirely, but likely reducing it.
It could turn out like the California “chemicals present that could cause cancer” that are ubiquitous in California, and therefore not useful as signal.
Speed limits work nonetheless. A few hefty fines and most of it will stop.
> It won't work.
Maybe. But the rest of your post doesn't have anything to back this up.
> I'm also curious, what about AI assisted media? If a company writes a full article, then uses AI to edit it a bit, does that need to be labeled?
Why would it? The threat is mass generation of articles without any human oversights, and realistic looking images claiming occurrence of events that never hapened. Nobody cares about an AI typo and stylecheck.
I would hope so.
Makes sense. Same as "Made in XXX" label. Users have the right to know where the product was made and how synthetic the product is.
Skip to the end. Let the consumer verify that the product is the unmodified intention of the entity claiming to have produced it and let them decide if they trust it.
Tangent: it is notable that certain variations of labelling of that sort would now be considered unacceptable by many:
https://web.archive.org.au/awa/20240617043532oe_/https://blo...
> Let the consumer verify
.. and ..
> and let them decide
Isn't it naive thinking ?
Not remotely. Let people check authenticity of the source and make their own private decisions.
The entire problem is fraud, and governments are hardly immune from that.
"Let the customer verify if the flour has asbestos in it, and decide if they trust it."
"Let the customer verify if the car airbags were installed properly, and decide if they trust it.
"Let the customer verify if the plant that produced the plastic in their grocery bags is dumping its waste products in the water nearby, and decide if they trust it."
No.
We have governments and regulations for a reason. It is utterly unreasonable to expect customers to verify everything themselves. We need clear, firm labeling laws for all kinds of products. There's no reason this should be different.
Every website will just add a global label somewhere in the pop-up they already have about cookies.
Label additions done by machines vs those by humans?
An obvious start would be that any such technology needs to have its own clearly visible watermark and removing it constitutes a crime.
Why and why would it need to be a crime? Again, we don't label machine additions, for example.
You will have to tell me how "labelling machine additions" is applicable at all as an example here. Unless your argument boils down to "computer doing things does not require labelling".
That is mostly my argument. I find "computer doing things" generally not a useful category and why here specifically doesn't have enough support. If there are some truly high risk gaps, then why no close them specifically?
EU once again at the frontier of AI.. regulation.
Hard to assess what this rule will actually accomplish. Ultimately AI will be involved in all creative processes going forward, so all pieces of media will have at least some basic AI disclaimer. At the same time, the bad actors that the EU is so paranoid about (Russians secretly undermining our democracy and other boogeymen) will obviously not use the label.
But it's another tool in the EU's toolkit to arbitrarily enforce fines on big tech when politically convenient, of course. That's one thing we remain good at.
I'm surprised by the sentiment.
With enough money you can already buy attention and presence. Now you can even facilitate a fake discourse to claim legitimacy and steer manipulation into entirely new realms. If this is not counteracted not only will rogue nations destroy democracies even faster, but powerful corporations and rich individuals will also use this to increase their power further and damage simple consumers.
Nothing about this concerns any creative process.
Rogue nations will not apply this AI label to the content they generate, so they will end up looking more authentic in comparison.
More authentic because of what, a missing watermark? It still has 2 positive outcomes:
1: It gives legal leverage against vectors of misinformation. Misinformation is tricky to combat legally and this is a first obvious measure that can be taken.
2: Reputation will become more important in the future, and if you are found to have repeatedly posted AI content claiming it is not it will damage you. Not to mention that users can point out the obvious discrepancy.
> Reputation will become more important in the future,
Citation needed
[dead]
> Ultimately AI will be involved in all creative processes going forward
What on earth makes you think that? Do you honestly think that literally all creatives will just start making lazy crap because… it’s cool? Because I can assure you, outside of this little corner of the internet, it very much is not.
Proper authors, journalists and artists are, generally, not keen.
There's creativity and there's the creative process. I like to dictate documentation in long form monologue, and then give instructions how to structure it. Then I give it a final pass with manual adjustments and the end-result is exactly how it would have looked if I had written it from scratch but without the bsywork.
Don't be stupid enough to let machines do your own thinking, treat them as a dishwasher.
Writing docs is, at best, only marginally creative work.
Though honestly, still probably skip the LLM if you’ve any pride in your work; LLM-‘assisted’ docs are generally worse than those written by people.
So what are journalistic texts if not docs? At this point, an LLM is just another evolution from the typewriter.
You are still responsible for your work. What is important are the semantics and statements of the text. Less so how the exact arrangement of letters came to be.
Journalists should not be using LLMs. Most proper papers ban them from using LLMs to write their stuff, for instance see https://www.rte.ie/news/2026/0319/1564281-mediahuis-peter-va...
You are not wrong, but here's the catch:
"Even I - with all my years of experience and knowledge - fell into the trap of hallucinations. I summarised reports using AI tools and worked from those summaries, trusting they were accurate.
He used an LLM to perform research instead of using it as a mere writing tool. I think large language models are great at formatting text, formatting, structuring and giving a stylecheck. But the last pass and review must always be a human. Everything posted by an LLM under your name is still under your name and you are liable for every single word. It's about the outcome and everybody who uses it is responsible for it.
> Proper authors, journalists and artists are, generally, not keen
The AI Scotsman fallacy.
In this context by proper authors I mean authors published by real publishers, for a start. Generally, most are hostile to LLM writing.
Also, like… why do you think people write books or make art? In the vast, vast majority of cases it is not to make money (virtually none of them do). It’s to _make_ something. Where’s the feeling of accomplishment in making them if you just told an LLM to shit something out? Like, who would bother?
Publishing is a business. There will be real publishers who are interested in using AI for the sake of making money. Even when a publisher has integrity and refuses to sell AI generated books, they have to contend with authors who are only interested in making money.
I'm not sure if I'm using the right term here, but publishers have been using rotating ghost writers to write series for generations and many of the authors have been happy to translate outlines into books just to earn a living. (I know of ghost written series that my father read as a child. I'm by no means young, and he had children late in life before it became fashionable to wait to have children until they were older.)
> In the vast, vast majority of cases it is not to make money (virtually none of them do).
If you're suggesting that virtually no authors write books to make money, I would suggest that is not true and the landscape is going to change for the worse. There are constraints on how many authors write books purely to make money because writing takes time and effort, with the returns being notoriously low. AI eliminates the time and effort factor. Returns don't have to be high if it is being generated by a content mill. If anything, diluting the market will result in lower returns for human authors -- discouraging them from writing in the first place.
> they have to contend with authors who are only interested in making money.
This isn’t _really_ a thing. If you’re into writing for the money, you’re probably in the wrong business. Even some quite successful/popular authors have a day job; only a very few can support themselves on it.
My apologies for a lack of clarity in that sentence, but I meant that publishers have to contend with people who are willing to use generative AI to make money by writing books for them.
That said, real people write for all sorts of reasons. Money is definitely one of them. Sometimes it is out of ignorance (e.g. they don't know that few books earn much money). Sometimes it is out of delusion (e.g. some authors are clearly raking in the money, they think they will too).
> Where’s the feeling of accomplishment in making them if you just told an LLM to shit something out? Like, who would bother?
Exactly. Find something else to do.
From this perspective the mere existence of AI tools for this stuff devalues all competing possible results, whether the tools are used or not, simply because there will be so much possible everything (media wise) on demand to match the whims of any consumer so they lose interest.
I’m pretty sure the big techs get safe harbor in this, otherwise they would have to stop serving up third party content in the EU all together.
Here's the actual law :
https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
I feel like this should be linked in any discussion of any legal matter
Surely a better idea would be to focus on how we can idk, get camera firmware to cryptographically sign photos or somethin, and then have editing software apply a signature on top. Or at least encourage a system where theres a good chain of image provenance for photojournalists
People have already tried this with C2PA, and it's not a perfect system. The main problem with these systems is that you risk certifying fake media, which is worse than uncertified fake and real media.
https://news.ycombinator.com/item?id=40313769
I do wish more people used basic watermarks on their media regardless - as in sloppy ffmpeg text. People are more likely to repurpose other people's older media than generate new media, and old time-y watermarks make that just slightly harder. Maybe a free app for uploading citizen journalism media that handles the watermarking with geo- and datemarking is the way to go.
I don't see how these are contradictory. The battle for reality will only worsen and you likely need multiple failsaves.
This is honestly not a bad idea. And could be done in parallel.
I even think that it might not need a mandate, if camera manufacturers start making it as a marketing claim. Prove your work is authentic and not AI generated!
Move comments to source? https://news.ycombinator.com/item?id=49124564
Cookie banner 2.0
This misses the forest for the trees. The important question is digital provenance so you can verify it came from the entity it claims; whether the entity you trust used AI is not the point. Plenty of people were not to be trusted before, only the EU would insist you trust and distrust who they say you should.
This is going to get very interesting, I could see some erroneously claiming something as AI and demanding that I have an AI label when it's not. I'm not sure how this is even enforceable.
What is the definition of "authentic looking?"
Law isn't code. Intention and context matters. You don't need a precise specification to mandate something. https://en.wikipedia.org/wiki/I_know_it_when_I_see_it
In general, in EU
- law doesn't play "wording nitpicking", "Simon said" or similar nonsense. And this applies here, too.
- law doesn't care about what you or any "specific" person thinks something means/looks.
i.e. if you ask 1000 random people if something "looks authentic" and more then a very very small fractions says it is "authentic looking" then it is authentic looking (idk. what "more then a very small fraction" is, but 10% is definitely more)
Honestly any honest interpretation of "do you try or thing a random person from the street might mistake it as authentic (based on the _looks_, not content)" is giving you a pretty good idea for what it means. (honest here == not definition nitpicking, not egocentric, not elitist, not defaultist, etc. )
But also most relevant:
_This (authentic looking) is the wording of the article not the law/regulation._
This is mostly about content which tries to manipulate people with deep fakes.
And stuff like it being clear if e.g. a AI backed voice translation was used, as there had been more then one or two cases of voice translations being wrong in a way you would attribute to malice for a human translator.
E.g. Ai generated voice overs putting
If you have issues with having to label your GenAI slop as "AI made", I'm going to assume you are not trustworthy. Just label it.
Another step to make users dumb and blindly trusting of their regulators. I don't opose labeling AI content, I opose mandating it and giving users the false confidence that something without label is legit.
Let’s stop labeling cigarettes are bad, because people will start smoking cigars and think they’re safe.
Unlabeled doesn’t mean safe or legit.
Let's start labling everything. Let us label every cigarette individually, every single drinking glass, every single nail.
No, people still smoke, drink and fornicate, despite all the labels.
At some point, we should just let Brother Darwin sort out the thinking ones from the non-thinking ones, because at a too aggressive level of protection, society will stop working, and people will regress to children.
Labels did work in reducing harmful behaviors:
> Participants drew on a range of risk discourses in framing their abstention or moderate drinking along the lines of health, wellness, wisdom and productivity. They reiterated social constructions of heavy or regular alcohol use as irresponsible, threatening and potentially addictive.
https://onlinelibrary.wiley.com/doi/10.1111/dar.13709
The non thinking ones have voting power too and they will burn this shithole down because they think some powerful people in nice suits act on good faith and should be given all the decisionmaking and resources.
This way of thinking is a very efficient way of destroying your culture and civilisation.
You suggesting the inverse?
Consciously labelling something as genuine / non AI incorrectly is fraud.
"Forgetting" to add an AI label is an error of omission.
Feels like the former is far more punishable than the latter.
If I buy a bottle of, say, white spirits, there’s a warning on it that it’s poison and dangerously flammable. Because that’s the law. Now, there’s nothing _really_ stopping an unscrupulous manufacturer from making a bottle without those warnings. But on balance it still feels preferable to force manufacturers to do it.
My biggest problem with this is that this regulation is only aimed at AI. It should be much broader and include any technology that can be used to make 'authentic-looking' content. Why should someone who's good at Photoshop or VFX be allowed to share fake stuff as real, but not someone who's good with AI?
For many years we've been aware that edited models on magazine covers, Instagram and in ads are negatively impacting the well-being of (young) people, but we've never done anything about that. I consider this regulation incredibly hypocritical and full on populist because 'AI BAD!' and nothing else.
In Norway ads already must specify if a person in it has been digitally altered (changed shape, smoothed skin etc). Quite funny with international ads getting these banners plastered over their face. Random example: https://imgur.com/a/t8zT37o (rules for size and contrast)
Here are the rules: https://www.forbrukertilsynet.no/vi-jobber-med/merking-av-re... Quite level headed and easy to follow. So no, editing contrast or whatever people think they're clever to point out doesn't count.
Where does it stop? Should it include edits with a digital camera? What about black and white?
I'm not trolling. I'm thinking about how you practically define what you are proposing to regulate.
> incredibly hypocritical
If that's all... If anything, it's lenient for AI. You can still spread all the mind poison you want, you just need to label it.
> full on populist
It's not populist. That's just a cheap frame.
AI is capable of generating things that are nearly indiscernible from reality, with almost any input, be it adversarial or other. Photoshop at least involved some manual trimming and lengthy process that couldn't be applied to "literally anything you can thing of".
See these:
- https://www.instagram.com/p/DTDKbhflVRO/ - https://www.bbc.com/news/articles/ckg8wvz427vo#:~:text=Accor...
And many, many examples in day to day reporting.
I'd rather have the laywork already there to punish sensationalist publishers that abuse AI to forge alternate realities than nothing.
Manipulating information is going to be the plague of the post-GPT/post-Midjourney era. We are entering an era where trust if going to be hard earned, and I don't know how you can trust for-profit or politician/billionaire owned companies anymore. What prevents WaPo from manufacturing information?
You completely disregard practical implications.
Can you manually create images that were done via AI? Of course you can.
Can you create AI fake content and misinformation as easily with human labor as you can with AI? No.
Understanding this distinction answers whether you see the implications AI has on the future of democracies or not.
What about the practical difficulty of defining "AI"? You could explicitly call out, say, diffusion models, but that would leave future holes for new AI technologies
What about Photoshop being driven by Claude Cowork? That my seem asinine, but it is a serious consideration
I found this definition here [1]
> ‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments;
But that barely clears anything up. Even worse, I would not include a simple user-prompted image generator under this definition, as it is not operating with any autonomy. Or, you interpret it so broadly that seam carving [2] is considered AI, let alone content-aware fill
[1] https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
[2] https://en.wikipedia.org/wiki/Seam_carving
"It is a matter not only of customer protection, it’s also a matter of democracy protection."
Ensuring democracy by selectively censoring. I feel safer already.
I better put AI label on every video I make as there is already AI in the supply chain of all the tech I use. All content contains AI now.
Let's talk about AI commissars who will decide which company will survive. In fact, let's create a few an NGOs to help! NGO will tell commissar what he wants to hear, NGO will get rewarded and allowed to exist!
I doubt EU will survive the next decade or two without major crisis or war tbh.
Edit: Instead of downvoting, how about you try retorting? I don't think I'm incorrect here. But maybe I am!
I am seeing more and more this exact sentiment that is in this comment: "I should be able to bamboozled everyone into thinking I have skills I actually don't have anymore".
Look, there's a ton of places where you can display your AI code, products or comments. Why is it so damn important that no places remain for people who want to be aware of whether or not they are engaging with a human?
Why is it so important to you that users must not be aware?
My suspicion is that it will quickly not matter in a lot 9f situations for most users, customers, etc. - so might be easier and better to curate places of true human interaction, for example, instead of the other way around.
The outcome will be that places insisting on human content will become more valuable, not less.
Even the most ardent AI booster wants their slop to be viewed by human eyeballs, ironically.
Slapping an explicit "I did not put effort into this" label on slop content is not censorship, its quite opposite, it lets people filter out the noise. No human can hope to be able to do this on their own at the scale GPUs can pump out tokens.
Authentic voices are today being drowned in waves upon waves of AI slop spam. This is not quite censorship but the end effect is similar - human voice cannot be heard. Just this time not because someone stopped the human from talking, but because they as shouting 1000x louder instead, at pennies on the dollar it costs to create something truly valuable.
Some generic picture used as like icon: is the labeling critical there?
Probably not critical, just helpful to be able to understand where the source is willing to cut corners. Helps me understand if the given content source values are aligned with mine without risking wasting time to ingest slop just to come to the same conclusion.
The same way I look at food labels to support local produce and those who don't try to trick me into thinking I am consuming something it's really not.
I personally wouldn't care for AI usage in icons, I don't put a lot of attention to them. Someone else might.
Buying a license for a stock photo vs using AI to create what you need is cutting corners? Don't thing there is much difference in effort (might even be more on the latter)?
Buying a license for a stock photo is paying someone who has put the effort.
Well, okay, I am not sure how much effort goes into creating your typical stock photo, probably not a lot, just still more than 0 effort it takes for the AI to generate it.
And yes, I see difference between an automated system flooding the web with AI generated "news" in disguise vs a niche blogger using AI to iterate on an image with hand crafted prompts. Labels are not perfect, just better than nothing.
Compliance burden tend to favor the large, which is not what I think is a direction that needs more pushing into.
(I also think that labels will quickly become just "background noise" that is mostly ignored.)
requiring deep fakes to be labeled as deep fakes has nothing to do with censorship
there is also nothing selectively about it, it applies to all deep fake equally
if you had read the article you would know that this is mainly about deep fakes and other deceptive content, not every use of an AI tool (e.g. line smoothing) requires an AI label, specifically it's only about content "generative AI" usage _in a context where it's potentially deceptive_.
> AI commissars who will decide which company will survive
a huge jump in logic not related to anything in this article
> I doubt EU will survive the next decade or two without major crisis or war tbh.
again unrelated, but any country which doesn't fights deep fake propaganda will most likely not survive as a democratic state of law, soo if anything this action increases the chance of survival
> Edit: Instead of downvoting, how about you try retorting?
- then maybe read the article properly, your comments put it highly in question you did
- also maybe don't have non logically sustained jumps in reasoning in your article, especially not when you jumps in reasoning come with tons of sarcasm
> I don't think I'm incorrect here. But maybe I am!
well yes, you have even missed the long known, explicitly pointed out sever threat deep fake propaganda is for any state and especially democracies. That is not a good bases for any argument.
Also passive aggressive edits are a good way to get even more down votes if anything.
Pros: none
Content label doesn't transform the content. It just provides a false sense of category.
I still have to think critically whether there is a label or not.
Cons: New ways for sociopaths for corruption, embezzlement and censorship.
That is my point.
This idea that legislation has no downsides is extremely naive and borderline retarded. Look at proposed legislation from the point of someone who wants to destroy everything. Does it provide a path to that goal? Is the risk worth it?
Every government grows over time and produces more and more suffering while consuming more and more wealth. I recommend reading The Gulag Archipelago. Your thinking seems to be dangerously naive to you and people around you.
Labelling stuff as “this was made by a magic robot” is not censorship. Don’t be silly.
I'm OK with it if no action is taken when content is not labeled and public officials work for free.
Label doesn't prove anything. It just provides a false sense of certainty.
what's the alternative? Let our benevolent techbro oligarch overlords decide what is best for us?
I think there is no alternative that yields good results.
Legacy media uses the argument that consumer already expects to be lied to and knows the content is for entertainment (TV news for example). I feel the same about AI content.
Or should I be expected to blindly trust everything? I already have to critically think and filter everything already so why the distinction? I should try harder for some content and not another? Especially as AI gets better the distinction between AI and no-AI generated content gets blurrier.
All I see is bureaucrats trying to grab more power over people's decisions and more money embezzled. Unelected people forcing rules on their subjects.
Nihilism is the path to death and destruction. Thinking this will makes you powerless and cedes control with those who have the most resources.
It's easy to take yourself out of this equation, but it only makes everything worse. The trick is that they make you think that it couldn't get worse. But this is always false.
Can you elaborate? I have hard time seeing nihilism here. (there may be language barrier and/or definition issue at play - maybe?)
With definition of nihilism along the lines of `the rejection of all religious and moral principles, in the belief that life is meaningless`
The point I am making (or trying to) is quite the opposite. Bureaucrats seem to be introducing more violence and even more problems while resolving none. I still have to critically about the content I consume regardless of it having some label. But now tax money goes towards this instead of the things citizens voted for.
You think that the status quo sucks and that there is no alternative, that's your prerogative. But why do you complain about people who think there is an alternative and try to implement it?
No issues with people thinking anything. Big issues with people using violence and death threats enforcing their image of the world though.
As long as it's privately or voluntarily funded and no violence or coercion is used, then I don't really care.