If you run, store, or fine-tune LLMs, the EU bureaucrats will come after you.
This is what you get when you pass overly vague laws that can be construed as applying to literally everything.
Thalidomide Vintage Ad Shirt $22.14 |
Ape Out Shirt $21.68 |
Thalidomide Vintage Ad Shirt $22.14 |
Left wing Germans are and will forever be the most high iq cucked gigachads on the planet. It started before Karl Marx nor will it end with this Austrian guy.
Based. Frick ~~*OpenAI*~~, down with AI israelites.
>t. I use the Brave Browser and share everything I do on my computer with Peter Thiel
>A police nanny-state who controls what you can run on your own computer is good as long as it harms the jooos as well
I doubt you understood me. Since you use the Brave Sir Browser you are nothing but a gaymer child who couldn't hide themselves anyway (since you're already sharing your life history with Peter Thiel). At that point worrying about some random LLM should be least of your concerns.
I don't use Brave,
and
are two different anons.
>botnet by guy with ethnicity x bad
>botnet by guy with ethnicity y good
Spbp
But the EU is a talmudic cult and the GDPR is satanic dark magic what harms everyone not just OpenAI.
>This is what you get when you pass overly vague laws that can be construed as applying to literally everything.
That was the point. The EU is a bureaucratic hellhole where a few extremely stupid, smug, powertripping sociopaths control the economy of a whole continent.
>be eu
>American call you socialist
>American also call you oligarchy
>realizing German and English Americans have brain damage and go back to insured work
There's nothing vague about GDPR. If someone processes your personal information, you have the right to know about it, know *why* they're fricking doing it, know *what* kind of data they store on you, and have the right to demand change or deletion of the data. And it has to be protected from third-party access, unless you're informed what third parties have access to it and for what purpose.
Just because OpenAI processes people's info among a big ass data magma stolen from God knows where, doesn't mean the law is vague or magically stops applying to them.
It's weird the article focuses on "factual accuracy" of the information though, because that seems to be the least important part of the problem.
Anyway, frick 'em. If their AI is so great, surely it can find an intelligent solution, right?
>There's nothing vague about GDPR. If someone processes your personal information, you have the right to know about it, know *why* they're fricking doing it, know *what* kind of data they store on you, and have the right to demand change or deletion of the data. And it has to be protected from third-party access, unless you're informed what third parties have access to it and for what purpose.
It's going to be really cool to see this used by politicians to shut down news articles about them. All the major news companies have websites, and they are "processing personal information" whenever they publish an article about political corruption. They have every right to ask for that data to be changed or deleted.
Mentioning a public figure's name in an article is not processing personal information in the GDPR sense, and politicians have been trying to shut down news articles about them for years now with all kinds of SLAPP bullshit lawsuits. Where have you been living?
>Anyway, frick 'em. If their AI is so great, surely it can find an intelligent solution, right?
lol All AI's hallucinate. The law basically outlaws all AI. EU becomes a fifth world shithole.
Intellectual property is a harmful misguided concept. A categoey error. Its also incompatible with technological development.
europoors can't help but contribute NOTHING for humanity
Excuse me? Europe is cradle of civilization.
That was long time ago. Now its a socialist system ready to collapse, full of degenerates.
>full of degenerates
UK isn't EU anymore.
I dont get the idea behind brexit when both sides implementing the same moronic globalist tyranny.
>feed it literal misinformation and fake news by yurinalists
>hurr durr we can't control false information and fake news
The main issue with this shit is that they'll never admit that the information they give the ai is false because that would mean actually doing their job before pressing the train button, plus employing people who care about providing factual information (which knowing they're Californians means they don't care as long as it benefits their political agenda)
You are not allowed to have an universal statistical model with finite state. It's illegal to provide wrong answer for ANY question. Our stupid rules are more important than allowing you to solve real world problems for your own responsibility.
>information stored
You morons, just fricking read or watch a yt video how LLMs work.
They have the right to access that 200 billion float numbers if the want to read them.
LLMs can't be generated without storing and processing raw data, though.
Do you have mild to severe brain damage? Serious question.
Pic related. Sure looks like it can store accurate information about identifiable living people, while also returning inaccurate or outdated info (they divorced years ago) with no distinction made between the two. It seems very reasonable to me that people have GDPR concerns about it.
the paint brush does not store the painting, even if it can be used to reproduce it. reproduction does not equal storage
The model was fed information about John Carmack and Anna Kang during training and the model's weights were adjusted accordingly. The information was therefore stored in the weights.
This method of storage is imperfect, meaning reproductions of it may be inaccurate. OpenAI knows this but chooses to operate their ChatGPT service anyway. Meta knows thus but releases Llama's weights anyway.
ANY dataset might be outdated/wrong/corrupt. This is why the GDPR says that inaccurate data must be updated.
>Personal data shall be [...] accurate and, where necessary, kept up to date; every reasonable step must be taken to ensure that personal data that are inaccurate, having regard to the purposes for which they are processed, are erased or rectified without delay
Doing so isn't technically feasible for current LLM designs. LLM creators/operators are simply unable to do it. They know this and continue to provide their datasets and services anyway.
>This is why the GDPR says that inaccurate data must be updated.
But it cannot be rewritten like in a text database. It takes millions of dollars to train a model that large.
GDPR is moronic.
But the EU shouldn't even have the right to make rules like the GDPR.
The EU does this because 90% of Europeans are cowards who don't have any morals or values and allow their politicians to do this.
>But it cannot be rewritten like in a text database. It takes millions of dollars
The law says that new buildings must have fireproof materials. I have designed and built a building that isn't fireproof, however it would cost millions of dollars to redesign it now. Therefore I should be allowed to ignore the law. Makes sense, right?
Maybe they should have thought about this BEFORE spending millions of dollars building a blob of linear algebra containing a weird lossy encoding of questionable information about countless people.
Code is speech, and Europe hates free speech.
inb4
>herp derp harmful code and hate speech are not free speech!
It stores information like your brain. Your memories can be outdated or wrong or you can remember incorrectly or fill blanks with generalization. Your brain doesn't store textual data.
The EU will regulate itself out of existence.
The EU is the Ebussy of politics
>everyone realizes that "AI" is crap
It's not crap. It's super useful. Just don't try to use a hammer for screws.
Also guberments shouldnt regulate AI. It's enough to punish theft, scams and violence. Everything else is personal responsibility.
>This is what you get when you pass overly vague laws that can be construed as applying to literally everything
It was painfully obvious it would come to this. This is the way the thing Eurogays describe as """""legislation""""" works. Of course, IQfy will pretend to such EU dick, while living in the only free country on earth.
Now imagine being at the beginning of a paradigm shift in technology. How do you handle this situation? Japan explicitly made it legal because they know this shit is going to be super important. The EU fricking banned it, lmao.
Frick OpenAI.
>b-but muh freedumbs
ChatGPT is not freedom. Fight for your real freedoms, not for your freedom to defend a stupid corporation that censors, controls, and keeps their models private so they can rent them to imbeciles, even though they didn't pay anyone for the data they use.
The GDPR is pure unfiltered cancer with a million terrible side effects it has, but fricking OpenAI could be the first time such asinine law does something right. Frick OpenAI.
the problem with all this is the goddamn AI label. people treating it like an intelligent machine when it's a revolutionary bullshit generator. sure, the bullshit gets pretty close, but it's goddamn impossible to get the "AI" to distinguish between what it "learned" and what it "hallucinated". What they treat like the devil - LLMs making shit up - is at the very nature of LLMs. They "learn" to make shit up, that's how it's able to do anything other than spit out what it was trained on verbatim and that's the whole bit that makes them remotely interesting. you can't separate this from 'dangerous AI misinformation', at least not with a transformer model
it will be used to go after open source models as well
Reminder that OpenAI should be called ClosedAI.
This. There is nothing Open about OpenAI. Pure proprietary corporatism
don't care. chuds need to kneel at the feet of openai. before them the field of lang ai was just a few fringe researchers. chuds need to fricking kneel and thank oai, without them they wouldn't have their garbage 1.1t/s localshit models hallucinating Black person because its a bad finetune from a random incel