this post was submitted on 08 Jan 2024
329 points (96.1% liked)
Technology
59651 readers
2683 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I'm not a huge fan of Microsoft or even OpenAI by any means, but all these lawsuits just seem so... lazy and greedy?
It isn't like ChatGPT is just spewing out the entirety of their works in a single chat. In that context, I fail to see how seeing snippets of said work returned in a Google summary is any different than ChatGPT or any other LLM doing the same.
Should OpenAI and other LLM creators use ethically sourced data in the future? Absolutely. They should've been doing so all along. But to me, these rich chumps like George R. R. Martin complaining that they felt their data was stolen without their knowledge and profited off of just feels a little ironic.
Welcome to the rest of the 6+ billion people on the Internet who've been spied on, data mined, and profited off of by large corps for the last two decades. Where's my god damn check? Maybe regulators should've put tougher laws and regulations in place long ago to protect all of us against this sort of shit, not just businesses and wealthy folk able to afford launching civil suits and shakey grounds. It's not like deep learning models are anything new.
Edit:
Already seeing people come in to defend these suits. I just see it like this: AI is a tool, much like a computer or a pencil are tools. You can use a computer to copyright infringe all day, just like a pencil can. To me, an AI is only going to be plagiarizing or infringing if you tell it to. How often does AI plagiarize without a user purposefully trying to get it to do so? That's a genuine question.
Regardless, the cat's out of the bag. Multiple LLMs are already out in the wild and more variations are made each week, and there's no way in hell they're all going to be reigned in. I'd rather AI not exist, personally, as I don't see protections coming for normal workers over the next decade or two against further evolutions of the technology. But, regardless, good luck to these companies fighting the new Pirate Bay-esque legal wars for the next couple of decades.
You are misrepresenting the issue. The issue here is not if a tool just happens to be able to be used for copyright infringement in the hands of a malicious entity. The issue here is whether LLM outputs are just derivative works of their training data. This is something you cannot compare to tools like pencils and pcs which are much more general purpose and which are not built on stole copyright works. Notice also how AI companies bring up "fair use" in their arguments. This means that they are not arguing that they are not using copryighted works without permission nor that the output of the LLM does not contain any copyrighted part of its training data (they can't do that because you can't trace the flow of data through an LLM), but rather that their use of the works is novel enough to be an exception. And that is a really shaky argument when their services are actually not novel at all. In fact they are designing services that are as close as possible to the services provided by the original work creators.
I disagree and I feel like you're equally misrepresenting the issue if I must be as well. LLMs can do far more than simply write stories. They can write stories, but that is just one capability among numerous. Can it write stories in the style of GRRM? I suppose, but honestly doesn't GRRM also borrow a lot of inspiration from other authors? Any writer claiming to be so unique that they aren't borrowing from other writers is full of shit.
I'm not a lawyer or legal expert, I'm just giving a layman's opinion on a topic. I hope Sam Altman and his merry band get nailed to the wall, I really do. It's going to be a clusterfuck of endless legal battles for the foreseeable future, especially now that OpenAI isn't even pretending to be nonprofit anymore.
This story is about a non-fiction work.
What is the purpose of a non-fiction work? It's to give the reader further knowledge on a subject.
Why does an LLM manufacturer train their model on a non-fiction work? To be able to act as a substitute source of the knowledge.
End result is that
So, not only have they stolen their work, they've stolen their income and reputation.
There's a big difference between borrowing inspiration and just using entire paragraphs of text or images wholesale. If GRRM uses entire paragraphs of JK Rowling with just the names changed and uses the same cover with a few different colors you have the same fight. LLM can do the first, but also does the second.
The "in the style of" is a different issue that's being debated, as style isn't protected by law. But apparently if you ask in the style of, the LLM can get lazy and produces parts of the (copyrighted) source material instead of something original.
Just as with the right query you could get a LLM to output a paragraph of copyrighted material, you can with the right query get Google to give you a link to copyrighted material. Does that make all search engines illegal?
Legally it's very different. One is a link, the other content. It's the same difference as pointing someone to the street where the dealers hang out or opening your coat and asking how many grams they want.
Websites that provide links to copyrighted material are illegal in the US. It's why torrent sites are taken down and need to be hosted in countries with different copyright laws .
So Google can be used to pirate but that's not it's intention. It requires careful queries to get Google to show pirate links. Making a tool that could be used for unintentional copyright violation illegal makes all search engines illegal.
It could even make all programming languages illegal. I could use C to write a program to add two numbers or to crawl the web and return illegal movies.
Oh. Linking and even downloading torrents is legal in my place. Hosting and sharing is not. My bad.
From how I understand it is that the copyright holders want the LLM to do at least the same as Google is doing against torrents: it checks so no parts of the source material is in the output.
What does this mean? I don't care what you (claim) your model "could" do, or what LLMs in general could do. What we've got are services trained on images that make images, services trained on code that write code etc. If AI companies want me to judge the AI as if that is the product, then let them give us all equal and unrestricted access to it. Then maybe I would entertain the "transformative use" argument. But what we actually get are very narrow services, where the AI just happens to be a tool used in the backend and not part of the end product the user receives.
Talking about "style" is misleading because "style" cannot be copyrighted. It's probably impractical to even define "style" in a legal context. But an LLM doesn't copy styles, it copies patterns, whatever they happen to be. Some patterns are copyrightable, eg a character name and description. And it's not obvious what is ok to copy and what isn't. Is a character's action copyrightable? It depends, is the action opening a door or is it throwing a magical ring into a volcano? If you tell a human to do something in the style of GRRM, they would try to match the medieval fantasy setting and the mood, but they would know to make their own characters and story arcs. The LLM will parrot anything with no distinction.
This is a false equivalence between how an LLM works and how a person works. The core ideas expressed here is that we should treat products and humans equivalently, and that how an LLM functions is basically how humans think. Both of these are objectively wrong.
For one, humans are living beings with feelings. The entire point of our legal system is to protect our rights. When we restrict human behavior it is justified because it protects others; at least that's the formal reasoning. We (mostly) judge people based on what they've done and not what we know they could do. This is not how we treat products and that makes sense. We regulate weapons because they could kill someone, but we only punish a person after they have committed a crime. Similarly a technology designed to copy can be regulated, whereas a person copying someone else's works could be (and often is) punished for it after it is proven that they did it. Even if you think that products and humans should be treated equally, it is a fact that our justice system doesn't work that way.
People also have many more functions and goals than an LLM. At this point it is important to remember that an LLM does literally one thing: for every word it writes it chooses the one that would "most likely" appear next based on its training data. I put "most likely" in quotes because it sounds like a form of prediction, but actually it is based on the occurrences of words in the training data only. It has nothing else to incorporate to its output, and it has no other need. It doesn't have ideas or a need to express them. An LLM can't build upon or meaningfully transform the works it copies, it's only trick is mixing together enough data to make it hard for you to determine the sources. That can make it sometimes look original but the math is clear, it is always trying to maximize the similarity to the training data, if you consider choosing the "most likely" word at every step to be a metric of similarity. Humans are generally not trying to maximize their works' similarity to other peoples' works. So when a creator is inspired by another creator's work, we don't automatically treat that as an infringement.
But even though comparing human behavior to LLM behavior is wrong, I'll give you an example to consider. Imagine that you write a story "in the style of GRRM". GRRM reads this and thinks that some of the similarities are a violation of his copyright so he sues you. So far it hasn't been determined that you've done something wrong. But you go to court and say the following:
How do you think the courts would view any similarities between your works? You basically confessed that anything that looks like a copy is definitely a copy. Are these characters with similar names and descriptions to GRRM's characters just a coincidence? Of course not, you just explained that you chose those names specifically because they appear in GRRM's works.
deleted
Its wild to me how so many people seem to have got it into their head that cheering for the IP laws that corporations fought so hard for is somehow left wing and sticking up for the little guy.
deleted
Just a heads-up, libertarian is usually understood, in the american sense, as meaning right libertarian, including so-called anarcho-capitalists. It's understood to mean people who believe that the right to own property is absolutely fundamental. Many libertarians don't believe in intellectual property but some do. Which is to say that in american parlance, the label "libertarian" would probably include you. Just FYI.
Also, I don't know what definition of "left" you are using, but it's not a common one. Left ideologies typically favor progress, including technological progress. They also tend to be critical of property, and (AFAIK universally) reject forms of property that allow people to draw unearned rents. They tend to side with the wider interests of the public over an individual's right to property. The grandfather comment is perfectly consistent with left ideology.
And your argument boils down to "Hitler was a vegetarian, all vegetarians are Fascists". IP laws are a huge stifle on human creativity designed to allow corporate entities to capture, control and milk innate human culture for profit. The fact that some times some corporate interests end up opposing them when it suits them does not change that.
deleted
I already have:
I thought that was a prima facie reason for why they are bad, And no I do not believe all copyright law is bad with no nuance, as you would have seen if you stalked deeper into my profile rather than just picking one that you thought you could have fun with.
deleted
There are plenty from people who actually study this stuff.
I don't have a significant opinion on the Disney case, though I will note that it stems from the fact that corporations are able to buy and sell rights to works as pieces of capital (in this case Disney buying it from Lucasfilm).
deleted
Stifling a writing tool because GRRM wants a payday, on the basis that it can spit out small parts of his work if you specifically ask it too, is the opposite of advancing the art.
...yet allowing individuals to build upon existing works. Its literally the rest of the statement you put in bold, stop trying not to see on purpose.
deleted
I'm clearly talking about the technology when I say tool (large language models) and not the company itself.
If we can't freely use copyrighted material to train, it completely and unequivocally kills any kind of open source or even small to medium model. Only a handful of companies would have enough data or funds to build LLMs. And since AI will be able to do virtually all desk jobs in the near future, it would guarantee Microsoft and Google owning the economy.
So no, I'm not taking the sides of the corporation. The corporations want more barriers and more laws, it kills competition and broadens their moat.
I don't think GRRM is evil, just a greedy asshole that's willingly playing into their hand. I also don't think loss of potential profit because the domain has been made even more competitive equals stealing. Nothing was stolen, the barrier for entry has been lowered.
This isn't helping anyone except big name author, the owners of publishing houses and Microsoft. The small time authors and artist arent getting a dime. Why should literally the rest of us get screwed so a couple of fat cats can have an other payday? How is this advancing the arts?
deleted
It doesn't matter what the subject is about, tim clearly not saying OpenAI the company when I use the term "writing tool"
I'm advocating for us and society as a whole. If only google and Microsoft hold the keys to AI, we all end up paying a surtax on everything we buy because every business will be forced into a subscription model to use it and stay competitive.
There is too much data involved to ask for consent, you would just end up with big players trading with each other. The small artists wouldn't get a dime, only Getty and Adobe. It's literally not pheasible.
Nothing was stolen except future potential jobs. You can't own a style or anything of the kind.
The small artists aren't going to get any kind of benefit out of these lawsuits. It sucks that it's even more saturated of a market but the good ones learn to use these tools (LLMs and img/vid gen) to elevate their own art and push the boundaries.
Sure. Trickle-down FTW.
If I want to be able to argue that having any copyleft stuff in the training dataset makes all the output copyleft -- and I do -- then I necessarily have to also side with the rich chumps as a matter of consistency. It's not ideal, but it can't be helped. ¯\_(ツ)_/¯
Wait. I first thought this was sarcasm. Is this sarcasm?
No. I really do think that all AI output should be required to be copyleft if there's any copyleft in the training dataset (edit for clarity: unless there's also something else with an incompatible license in it, in which case the output isn't usable at all -- but protecting copyleft is the part I care about).
Huh. Obviously, you don't believe that a copyleft license should trump other licenses (or lack thereof). So, what are you hoping this to achieve?
I'm not sure what you mean. No licenses "trump" any other license; that's not how it works. You can only make something that's a derivative work of multiple differently-licensed things if the terms of all the licenses allow it, something the FSF calls "compatibility." Obviously, a proprietary license can never be compatible with a copyleft one, so what I'm hoping to achieve is a ruling that says any AI whose training dataset included both copyleft and proprietary items has completely legally-unusable output. (And also that any AI whose training dataset includes copyleft items along with permissively-licensed and public domain ones must have its output be copyleft.)
Yes, but what do you hope to achieve by that?
In your mind are the publishers the rich chumps, or Microsoft?
For copyleft to work, copyright needs to be strong.
I was just repeating the language the parent commenter used (probably should've quoted it in retrospect). In this case, "rich chumps" are George R.R. Martin and other authors suing Microsoft.
Just because it was available for the public internet doesn't mean it was available legally. Google has a way to remove it from their index when asked, while it seems that OpenAI has no way to do so (or will to do so).
deleted
I hear those kinds of arguments a lot, though usually from the exact same people who claimed nobody would be convicted of fraud for NFT and crypto scams when those were at their peak. The days of the wild west internet are long over.
Theft in the digital space is a very real thing in the eyes of the law, especially when it comes to copyright infringement. It‘s wild to me how many people seem to think Microsoft will just get a freebie here because they helped pioneering a new technology for personal gain. Copyright holders have a very real case here and I‘d argue even a strong one.
Even using user data (that they own legally) for machine learning could get them into trouble in some parts of the developed world because users 10 years ago couldn‘t anticipate it could be used that way and not give their full consent for that.
deleted
Personally, I think public info is fair game - consent or not, it's public. They're not sharing the source material, and the goal was never plagiarism. There was a period where it became coherent enough to get very close to plagiarism, but it's been moving past that phase very quickly
Microsoft, especially with how they scraped private GitHub repos (and the things I'm sure Google and Facebook just haven't gotten caught doing with private data) is way over the line for me. But I see that more as being bad stewards of private data - they shouldn't be looking at it, their AI shouldn't be looking at it, the public shouldn't be able to see it, and they probably failed on all counts
Granted, I think copyright is a bullshit system. Normal people don't get any protection, because you need to pay to play. Being unable to defend it means you lose it, and in most situations you're going to spend way more on legal costs than you could possibly get back.
I also think the most important thing is that this tech is spread everywhere, because we can't have one group in charge of the miracle technology... It's too powerful.
Google has all the data they could need, they've bullied the web into submission... They don't have to worry about copyright, they control the largest ad network and dominate search (at least for now).
It sucks that you can take any artist's visual work, and fine tune a network to replicate endless rough facsimile in a few days. I genuinely get how that must feel violating.
But they're going to be screwed when the corporate work dries up for a much cheaper option, and they're going to have to deal with the flood of AI work... Copyright won't help them, it's too late for it to even slow it down
If companies did something wrong, have it out in court. My concern is that they're going to pass laws on this that claim it's for the artists, but effectively gatekeep AI to tech giants
Where, for example?
The European Union, for example.
That's not right. It explicitly is legal in the EU.
That is not how the EU works. Member states can get together to tarif and sanction behavior, but just because the EU generally allows something doesn't mean all member states have to abide. Different constitutions and all. Besides I'd like to know where exactly any EU resolution explicitly allows corporations to throw any data they have at any technology or LLM's specifically even when nobody ever gave consent to that. Corporations have to be quite specific for how they process your data and broadly saying "machine learning stuff" 10 years ago isn't really water proof.
No. EU legislation often has so-called opening clauses that allow member states to tune "EU laws" to their needs but it's not the default behavior.
You seem to have the GDPR in mind. It regulates personal data, meaning data that can be tied to a person. If that is not possible, the GDPR has no objections.