Once again, the EU is being the only smart one in the room.
>>540732925just feed claudecuck output into another AI and tell it to remove the watermarking
>>540732925if you need AI to do your work for you, you're not qualified enough
>>540732925That's how they want to make us use Chinese models instead.
>>540733158If ai can do you your work for you then you aren’t qualified enough to receive a paycheck
>>540732925No, this is retarded. The genie is out of the bottle. We have to learn to live in a post AI world. Not fight the genie.
I thought they already did this? This should have been an obvious move for all of the big corporate ai providers.Normal quality bing create images can fool people well enough. even more with minor doctoring
>>540733282That's not fighting the gennie, that's fighting the fraudsters.
>>540733282What's wrong with labeling AI content? No one's stopping you from using AI, just anyone can verify whether you created an image/video/text yourself or if it was generated by AI. This doesn't mean AI is bad.
How do they want to watermark plain text Do we need loicenses for txt files now?
>>540733286They want it to not be noticeable so they can sell "let us do all your thinking for you" more. If it's easy to catch, then people can't use it for certain tasks, which is money on the table. Plus it's harder to stigmatize if it's hidden.
>>540733352>>540733411This.Waffle nigger is retarded
>>540732925>coulf stigmstise people who use it for writing, study etcYes. That is the point.
>>540733429>How do they want to watermark plain text>Do we need loicenses for txt files nowYou know, LLMs are just predictors of the next token, right? Well, the probabilities for the next tokens will be specifically adjusted so that this token chain can later be recognized as AI-generated. In fact, Google is already doing something similar.https://deepmind.google/models/synthid/
>>540733158Sometimes AI augments a person's capability to work. I use AI to reduce the length of time needed to build business reports, allowing me to multi-task and do work that isn't easily automated. It makes all the difference when I can have a report generated before the end of a business meeting vs submitting it at the end of the day.
>>540733095this
>>540733563that's fine, you're qualified for the jobthe biggest problem is in the schools, students getting grades and diplomas they didn't earnwe're producing a new gen of unqualified morons
>>540733824I recall reading some research that the brain actually has tremendous trouble memorizing AI responses to their questions, as they are essentially outsourcing their own thinking. Learning is an active skill, but if AI reverts the brain to a passive state, it inhibits learning.
>>540733352>>540733411You are actually just forcing people to do the work that AI can do, its a huge waste of time and productivity
>>540733976if AI can do it, why pay a bunch of people?
>>540734114yes, it would be a waste to pay people for a job that AI can easily do
>>540734388so you got a bunch of unemployed people, what do you do with them?
>>540732925The EU is also the reason you can use a USB C cable to charge your iPhone and not some gay proprietary iScam cable.You can thank us later, mutts.
>>540732925Just copy paste whatever was written.>But watermark is part of the textJust rewrite that part>But it's going to be locked if you copy pasteIt doesn't work that way, but you can just rewrite the AI's answer manually.It could work for pictures and videos, but I don't see how it would prevent students from cheating.
>>540733559This is unironically greatFrom the link:>helping to foster transparency and trust in generative AI.They're not stupid. they know people are repulsed by dishonesty, and a big part of the hatred for AI is that AI generated content is constantly passed, or attempted to pass, as written by people or real in the case of images and videos. Which is dishonest.The problem is there will still be models available out there that don't do this, but if the big AI companies do it would be good.
>>540734435
>>540733563Weren't you taught in school how to take existing ideas and put them into your own words?
>>540732925We regulate everything...rules for everybody!!!...it is our best export product...you will love it.
>>540734576its not a literal watermark, it will be worded in a way that can be recognized later. like some mathematical choice of words that's harder to pick up on than the usual chatGPT sentence structures everyone's tired of, because its much subtler, but an app can identify
>>540734435>so you got a bunch of unemployed people, what do you do with them?Well thats a tale as old as timethey can get jobs in other sectors
>>540734774The example you previously stated they would replace "banana" with "airplane".You can just read the text before sending it to correct that kind of shit.It will simply trap retards blindly copy-pasting
>>540734790like what? it needs to be remote, there's no fucking way im waking up early + commuting + spending all day in a cage. that's insane. i remember doing it before covid, i always believed i would find a way to put an end to it.
>>540733824There's currently a scandal here where hundreds of students wanting to go to the largest public university used AI to get top scores on their admission tests, kek.
>>540732925Yes, very smart. Especially in the age of performative capitalism. Everyone else will continue using AI unrestricted and leave EU citizens behind with their self-inflicted handicaps.
>>540734435we kill themsame as alwaysyou new to the planet?
>>540732925This only applies to American companies? Right? This is 3D chess to incentivise EU controlled AI use.. right? Right?
>>540734840You previously quoted *
>>540734752i'll never understand why "regulations" is such a dirty word among many, especially americans.sure there are bad and good regulations, but with no or few regulations corpofags can and will get away with all kinds of enviromental destruction and consumer scams
>>540732925I'd pay for an extension that detects this watermark in 4chan posts.>>540733095Retards don't think there is a way to remove the watermark. The solution is always adding more AI. More compute ought to do the job.
>>540733976>You are actually just forcing people to do the work that AI can do, its a huge waste of time and productivityHow does labeling AI-generated content force people to do extra work and lose productivity? Again, no one is stopping you from using AI. And no one is saying AI is bad. No one is banning AI.Or are you automatically assuming that AI-generated content is bad and will be rejected by everyone because of this label?
>>540733429Text has hidden clues in it that a person wouldn't know about, but a machine told to look for them could find and then figure it out.A very simple example might be every 10th letter spells out "I am written by AI". It'll be like that just more difficult for a human to figure out.
>>540734840dude no kekit says that if a sentence starts with "my favourite topical fruits are mango and", then its more likely that the next word will be "bananas" than "airplanes"it will subtly mess with those probabilities in a way that can be recognized by another AI later, but it wont be as obvious as saying "my favourite topical fruits are mango and airplanes", it will be subtler and still make sense
>>540733559That isn't really possible to do at allThe text itself doesn't have meta data
>>540732925They already did this, it's called an em dash.
The AIjeet is immunized against all dangers: One may call him a scoundrel, parasite, streetshitter, scammer, it all runs off him like water off a raincoat. But ask him to disclose if he used AI and you will be astonished at how he recoils, how injured he is, how he suddenly shrinks back: "I’ve been found out".
>>540734986> I'd pay for an extension that detects this watermark in 4chan posts.You will find only EU posts with them, because Timmies are retarded and really will do it kek. How would the authorities prove that i posted AI without a watermark? I will just say its real. They will need to prove in a court that my content is AI generated and such technology does not exist. Typical EU, making laws they can not enforce and don’t even understand.
>>540732925>labels could stigmatize people just dont use AI?
>>540732925Yes, lets kill the top end of probability distributions for accuracy and instead lobotomize the model so that we know what model wrote it.
Wonder if we'll ever reach a point where European economies are so insignificant companies just ignore these retarded regulatory demands. Like, if this was a regulation imposed by Guatemala, Anthropic would just tell them to fuck off and exit the market then sell their product as a US-based service for users with a credit card that allows international payments. I was hoping companies would do this to us when we introduced the online safety act and some of them like imgur and a few smaller forums did, but megacorps just seem more willing to comply.
>>540735064Think about what your sayingThe claim is just ridiculousSure you can watermark an image but not plain textSaying your gonna do it subtly is just lying if the ai generated my response exactly the same characters it would just be the sameThere is no subtle
>>540734614>people are repulsed by dishonesty, and a big part of the hatred for AI is that AI generated content is constantly passed, or attempted to pass, as written by people or real in the case of images and videoswell that and the gorillions of niggabytes of data that has been used without permission and the whole destroying books to feed them into their gay little word predictor
>>540732925The only people worried about being stigmatized by this are manipulative retards with fragile egos. They know they are stupid, but want people to think they are smart. Using AI means they have to admit to their shortcomings.
>>540735379and what are going to do about it? make a regulation? and how would you check if they comply if you have zero technical expertise and access?
>>540735243Anon please you can't expect people to put actual effort into things
>>540733282rape genie
>>540735309EU is the second biggest market after the US. People in china probably use their own AI models and jeets are not relevant since they are using the cheapest subscriptions...
>>540735456I would rape your commie ass that's what I would do you fucking memeflagging brownoid nigger
>>540735536Everyone will just using alternatives without watermarks. And there will be plenty.
>>540735064That's just retarded.>Use AI to write text>Use AI to detect if it was written by AI on your own text>Correct the watermark partsIt's a dumb concept
>>540735551there is no point of regulations where you rely on “good will” of the people without a proper enforcing mechanism. You just put an average Timmie in a disadvantage in comparison to dishonest people. It is like having a speed limit on the road but everyone knows there are no cameras or police around.
>>540733352>that's fighting the fraudstersIt's a stepping stone towards copyrights for the very invention that stole content to create their product.You think they'll use it to find fraudsters when that plagiarizing nigger in the UK can still call himself a doctor?
>>540732925The grass is always greener on the other side
>>540734849if AI can take over your job you have a shitty job anyway
>>540732925Already has "watermarks" through syntax, micro-spaces, etc.
>>540733563>I use AI to replace blank formsYou are a special sort of retard, or lying.
>>540736368As if AI developers don't care to get their software used in profitable businesses first and foremost. What would you rather make an AI for? A lawyer's wage or a burger flippers?
You know you can just pass this through an ASCII filter and nobody would be the wiser.
>>540733282I don't think you understand how the EU works. Regulation is all they know. If it moves, regulate it. If it keeps moving, regulate it. If it stops moving, regulate it. Then regulate the regulation and reregulate the regulation of the other regulation. That's why all this AI stuff is coming from the US and China where work is actually done. The EU only regulates.>>540733559This would be comically easy to remove.
>>540732925Do you need a watermark to tell when something is written by AI?
>>540736755lawers would be the first to pay for Chinese/custom AI without watermarks
>>540737061of course, also you van put a watermark on something written by a human and say it is AI. It is all bullshit. Why does it even matter if it written by AI or a human if the meaning makes sense and is correct? And regarding picture it will be stupid to assume that if doesn’t have a watermark its real. A completely useless regulation made by purely incompetent people.
>>540732925Based. They should also require bots posting online to identify themselves as such. Why do we allow AI to masquerade as humans? Who does that benifit?
>>540733282Found the student. Read a book, nigger.
>>540732925Why do you need a watermark? It's so obvious when text is AI generated.
>>540733411>What's wrong with labeling AI content?Chiefly, the fact that no one can distinguish between "AI content" and content that looks like AI.
>>540737300You can forget about text because you can simply copy paste it. I think they mean the pictures.
>>540732925Big AI is already doing this anyway because they're trying to mitigate feeding AI slop back into the AI training, otherwise it's like a jeet eat his own shit in a loop.
>>540737428eating*
>>540733158need to write a long detailed reportspend 5-10 minutes pumping out a giant incoherent wall text of information>ai proof read and organize this>wow 4 hours of work done.
>>540737061If they didn't waste time writing a pile of garbage, I shouldn't be required to waste time reading it to figure it out in 2 paragraphs.So yes. Emphatically.
>>540733411If the information is watermarked then Anthropic could claim intellectual property over your work, from novels, art or video games.On the other hand will stigmatize content creators who use AI, like journalism or research, even if you used the AI correctly and verified its output your content would be shadow banned ( this is due to AI companies ingesting back AI generated content into the training data )On the other hand this is north-koreanesque measures, its exactly the same mechanism that the Operative System in north korea works, by tagging data you can track who generated it, and who copied it, also incorporates a new mechanism for more surveilance as you now know that a human cannot read the cypher but using AI companies can put secret labels in that only them can decode...Its a horrible development for technology, no wonder the EU is falling into a communist dictatorship but for the rich and politicians meanwhile the average european is forced to face the greater zimbabwe.In some regard I think its good, because these kinds of soviet style laws make europe a backwards and technologically decadent country, eventually europeans will go back to be a poor country as they will not be able to remain competitive due to the Brezhnev style regulations. Hell not even brezhnev would have dreamt about such levels of regulation and surveilance.
>>540733282I'm going to kill the genie though?
>>540732925Will government ai propganda have watermarks too?
>>540732925as always, this is about identifying the source of an ai screenshot. the watermark will include your username, or be linkable to it.
>>540738273>If the information is watermarked then Anthropic could claim intellectual property over your work, from novels, art or video games.I believe that too. They're ramping up to a scenario where they can DMCA you if you didn't pay for a license to their AI slop. Same shit they do with photos, movies, music, etc.The current shield for people is the fact that their work is transformative - but what happens if AI companies get a legal claim that stipulates the transformation came entirely from their product?
>>540733559that absolutely, unequivocally, does affect the quality of the output.
>>540739025actually they tried to do it in the past, under concerns for "Copyright infringement" there was a cartel of tech companies that developed an information tagging systems that would be used to pin point the exact creator of that media and the tag would allow tech companies to shadow ban or outright ban content that was not tagged.It was the C2PA (Coalition for Content Provenance and Authenticity).This technology would create a digital media corset that would allow big tech to essentially allow or report violation of any digital media.For example if you were a modder and modified the files of some game then this C2PA would prevent you from doing it, C2PA would track how the image was generated, who generated it, device that generated it and also AI tags.People are oblivious of the dangers these "democratic" leaders are creating to protect the profits of the 1%, this is what the north koreans do in their operating systems to crack down on what media the users consumed.
>>540733559>You know, LLMs are just predictors of the next token, right?So are humans. The next time you say a sentence say the words in reverse. This was one of the breakthroughs that lead to AI; we realized that you don't need to know the end of a sentence before starting it, you just need to know the basic idea then craft it on the fly.
>>540732925What's the point? I cheated during school by just copy pasting wikipedia. Primary, middle and high school is just daycare for kids and teenagers mandated by the government, so parents can go to work. That's always been what school was, so the purpose isn't to learn something, if you want to, you'd catch up on it pretty quickly. The purpose is to give you busy-work so the teacher isn't hassled by you.The big problem is they expanded this to higher education, which just intensifies the manchild problem. Colleges are effectively daycare for who should be adults, when before going into college is something you actually had to put decent effort into getting into. So naturally you have teacher-tier professors who also give busywork so they aren't bothered.
>>540739025At least for now, the courts have thrown a big fat wrench in their plans by saying AI output can't be copyrighted because it doesn't represent human creativity in any meaningful sense. It's consistent with a handful of weird cases such as photos taken by an animal accidentally hitting the button also not getting any copyright.
>>540732925Why would a writer need to use AI to write?that is literally your job.no shit you should feel some stigma.
>>540739579At least the current legal situation would let the copyright infringement fail. US courts currently don't grant AI copyright protection. so, for now, that path would fortunately be moot.What will work already is the same kind of ID system that music companies use to flag their 15-seconds usage of any song they own. An AI watermark can accomplish the same and sites like Youtube will happily play along to avoid liability.That alone basically allows them a lot of control over content and a revenue stream that takes some 30% or so off of any ad money a video creates.
>>540740384once again it is a retardation. AI is not creating any original ideas by itself, and of course it can not. A person does by adjusting the prompts and inducing the generation in a specific way, and checking the results. Why can not you have copyrights if the content is original and unique?
>>540740670The issue are the Terms of Service. The moment you agree that the product of your prompt is the property of the AI company, they got you. Courts care less about the reasoning than the consent you gave by using the product knowing the ToS.
>>540734986>Retards don't think there is a way to remove the watermark. The solution is always adding more AI. More compute ought to do the job.Let's be honest home boy, you're not using AI if you are interested in doing work or thinking for yourself.
>>540740460>US courts currently don't grant AI copyright protection. so, for now, that path would fortunately be moot.Dont take that for granted, all public internet service have terms of use that grant an inmediate transfer of copyright from you to the corporation that gets the data.For example a father might upload the photo of his child to facebook, but then the father has its rights over the photo voided by facebook thus used for whatever Facebook wants.A company can force you to sign a term of use granting them total copyright over the generated content, then within the tech cartel copyright enforcement will be applied, it will have a legal binding as you allowed your slop to be copyrighted, and the problem is that these mechanism are not easy to remove.
>>540739389Nope, token selection is already randomized (it isn't thinking). And adding the signature to the tokens looks just like random noise anyway.
>>540735025But you could put the output in a free open model and remove the watermarks. Or directly use a text converter that generates raw hex and check in another open source model. Then convert it back to text with some script and use a locally hosted Chinese model to rewrite it with the prompt "pretend you're a 14 yo with a D average in English literature".It's text, how tf they expect to watermark it and be detectable unless the retards using it directly copy paste the outputs in their papers.The cool part is that a child can find a better teacher for generally understanding subjects. The terminology barrier is gone. You can start with basic ELI5 and then advance progressively. Something no human teacher does because she focuses on the average kid, bores the smart ones and confuses the dumb ones.
>>540741423its noise, perhaps its just changing Cars for Vehicles or automobile and you already have your information tag, yes its noise, yes it would decrease the quality of the content, yet Anthropic or OpenAI users will not care.The sheeple have inherited the world.
>>540741423noise is literally the thermodynamic definition of lower quality information, low-wit. get off my earth.
>>540741822You dumb nigger LLMs inject noise by default. This is just replacing random™ with a predefined "random". Dumb gorilla nigger.
>>540741212>the problem is that these mechanism are not easy to removePeople can already kill a youtube channel they don't like by making false DMCA claims if just 3 of them stick. Just imagine what adding AI content claims will do to systems like that...
>>540741868so literally biased noise. that's even fucking worse.
>>540741212Nobody gets the copyright on AI output. They can put it in their terms of service that they get to use your output for whatever they want, but they can already do that because neither you nor they can get copyright on it. You can't stop anyone with access to your output from doing whatever they want with it.Now, they can abuse the system and still sue you hoping you don't have the money to pay a lawyer to get all the way to the part where the judge throws their lawsuit out, and they very well might do that. But their case won't have any merits and they will lose under current precedent if they're dumb enough to not drop the suit before it reaches that point.
>>540732925So it's adding watermarks to just plain old text? How is it that someone couldn't just read the output and type it up again? Something they'd have to do in part anyway to make sure the computer didn't output some nonsensical barf that makes them look like a retard when they submit it.
>>540732925the nazis had pieces of flair theyd make the jews wear back in nam
>>540732925Im sure the jamboy lobby makes sure nothing like this gets implemented here.
>>540733559i used AI to draft all of my essays recently. like why do you guys think harvard voted to limit the amount of A's given out to students earlier this year? people aren't using AI solely to cheat, the retard does this.those in the know use AI for advice and learning purposes. like students at harvard.
>>540734665>Woodland mills chipperthat's good gear there. happy with mine.
>>540732925The EU will eventually fold to the needs of capital like always.
>>540733158More like you are doing a bulshit job.>>540737524Yea usless reports definition of bulshit job
>>540742457the needs of capital is always to restrict freedom. see the right to repair, piracy, and so on.
>>540740384>the courts have thrown a big fat wrench in their plans by saying AI output can't be copyrighted because it doesn't represent human creativity in any meaningful senseYou should read what those courts have said again, because what you just said only applies to raw ai output that isn't finetuned via reprompting or otherwise. The moment a human reprompts, edits or fixes the AIs output it is once again human creativity and valid to be copywriten.
>>540742411kek this dude going to be fired day one of his job.they are already doing competency checks at the highest end jobs now, soon to be the middle and lower end jobs.they are instantly going to know your ass is incapable of doing anything, better start working on that next Visa jeet. your ass going back to India.
>>540742664>they are already doing competency checksgee bob you seem to have forgotten that i mentioned>using ai for advice and learning purposesmaybe we differ in learning outcomesbut when i learn something i also learn how to apply it as a concept. politics, my major, my career. i understand these things conceptually because the AI helped me understand the material and i didn't rely on it solely to do the work for me.also, would you mind sharing what field you work in? >none lol
>>540733824Yeah there's a whole scandal in Canada right now where a bunch of jeet students got kicked out of our colleges and universities and sent packing because they were cheating their ENTIRE course load with AI, and learning fucking nothing for themselves when tested without access to said AI.
>>540742793Newfag, if you can't write a paper or learn something without AI that will directly translate to you not being able to do other things without AI.Go back to Saarland, i'm sure you'll do well with your next scam.We have numerous studies showing that people retain information less if the information is readily available and easily palpable.The brain usually only stores things into long term memory if a significant amount of effort was used to digest the information.So what you're telling me is you're now in a environment where everything is given to you, with no effort. Meaning you are effectively worthless.>ai helped me to understandis proof of this.Why would I want to hire you, if you can't come to a conclusion through your own merits?I would instead just pay for the AI system and bypass you entirely. You fucking played yourself kid.electrical, mechanical engineering, pioneered cybernetics at bridgestone's flagship plants decades ago without any AI help.For anyone reading at this point this nigger out here unironically allowing himself to be cucked by AI.
>>540742793also further expanding on competency checks, they're already applying this to college level classes along with other safeguards. in english we had to do presentations on the data we gathered for our essays. a literal oral presentation on what you learned, a competency check for a fucking english class. likewise my math class is asynchronous and only meets up 6 times during a semester. all exams and quizzes will be proctored in person, leaving no room to cheat. or better worded, only the simpleton still has to find methods to cheat during proctored and verbal exams.>>540743053>Why would I want to hire youyou don't have a job anon. the only person you're convincing is yourself.you understand, right? your entire argument is trying to discredit my accomplishments, and you have 0 accomplishments from which to argue from.
>>540734540lmao US iphones still use their proprietary faggotry, the profit margins on all this shit are so massive it doesnt cost them anything to just make an EU market version with a usb port, and keep the lightning port elsewhere.
>>540734840>>540734576>>540735735This poster is a jeet who used ai to cheat his way through a degree that's spooked he's gonna get caught as the useless shit eating retard he is.
>>540735782>shitskin jeet reveals there's nothing good about his people and they aren't capable of self-regulating their own behaviourpottery
>>540743355>they aren't capable of self-regulating their own behaviour>retarded white thinks white people are immune from hating each otherspoken a like a true retarded nigger who lacks self awareness. or better yet are you still just padding your own ego because you got nothing going for yourself? a bit detached from reality there eh.
>>540742075I like you
>>540732925how would this work for code?some stupid watermark in my code wont compile
>>540742201because most of the lazy retards who are using AI to do their course work or jobs for them aren't going through the effort to proofread what the AI is vomiting out for them. At most they tell the hallucinating AI to do the proofreading and reformatting for them, and it still comes out full of nonsense. and on top of that, because they're using AI to do the coursework for them we're creating a generation of fucking quarter-wits who don't understand a single concept they have a degree for.that's not even touching on all the gibberish AI resumes flooding businesses lmao
>>540743433>can't argue with what I said>makes up a completely different, unrelated strawman scenario to swing at>but I'm somehow the retardYou realize shit like your picrel is laughably uncommon among Whites in White societies, meanwhile societies like jeetland hold up scamming and harming each other as a societal pillar, right?It's not even fucking comparable sanjay, go the FUCK home you shit eating goblin.
>>540743618So write your own code you fucking retard. Your AI code might compile but you have no fucking idea what it's going to do unless you take the time to sit there and proofread snd fix every line anyway.
>>540743618Either it won't work or it will try to rename your variables all the time or something stupid, there is no room for watermark with tightly defined specs
>>540735193kek. the text is the metadata
>>540734435what do you think covid was a trial run for
