OpenAI says it's "impossible" to create AI models without copyrighted material
arstechnica.com
arstechnica.com
From 2016:
> They’d raised a billion dollars and hired an impressive team of thirty researchers—but what for? “There are twenty to thirty people in the field, including Nick Bostrom and the Wikipedia article,” Amodei said, “who are saying that the goal of OpenAI is to build a friendly A.I. and then release its source code into the world.”
> “We don’t plan to release all of our source code,” Altman said. “But let’s please not try to correct that. That usually only makes it worse.”
Source: https://www.newyorker.com/magazine/2016/10/10/sam-altmans-ma...
have you tried saying the name out loud?
Was that a deliberate misdirection? Not sure. Apparently, Elon Musk takes credit for the name and wanted the company's products to be open source. However, Musk has said a lot of things about companies he is involved with that later turn out to be untrue.
Funny how putting forth lots of contrarian red tape after and because someone has made all the claims and promises has a way of doing that.
Imagine that Acme Music hires large sweatshops of workers, who are tasked with listening to trending (copyrighted) songs and then go through grueling practice to sing and play near-perfect performances of those songs, delivered to any customer who asks for one.
That business model is still copyright infringement! It remains so no matter how much you de-automate by having human-brains replace machines at the same tedious tasks.
P.S.: Kind of like the inverse of patents, where "$task, but this time with a computer" is not automatically a new patent-able thing.
whereas visually copying style somehow is more possible and ok
But we’re really talking about style. Think reggae, isn’t it immediately obvious what style of music you’re listening to?
Of course this is a super bad n=1 qualitative study with very bad response coding, but maybe this is simply because I haven't spent enough time with musical style copying LLMs (if there are any), and only spent about a minute listening to that forever death metal youtube AI channel.
Style is not subject to copyright, only expression. If there are unique lyrics and unique melodies under two bars then there is no copyright infringement.
A song that sounds like other pop songs is basically the very nature of pop music.
Pick a genre, for example reggae, and you’ll find a lot of perfectly legal songs that are almost indistinguishable aside from the musicians who are playing the music.
In this case you can assume Acme Music is making purposeful performances/impersonations of the original songs/artists, because the goal of the scenario is to show that "but I used a human brain to do it" isn't a magical get-out-of-infringement-free card.
Illustrating the boundaries of fair-use or transformative works is... Well, a somewhat-separate issue that would need a different kind of scenario.
Again, I am honestly confused by what is meant in this thread!
1. Conasg: Taking huge amounts of copyrighted material without consent and selling access to it seems illegal.
2. Intrasight: Humans learn from copyrighted material all the time, and indirectly profit from it, so why can't the AI?
3. Me: Some of the stuff the software is being used to do would still copyright infringement even if you had teams of humans (inefficiently) doing it instead.
There literally are teams of people paid to deliver reggae music to people. Or pop music. I’m not well versed enough in trap music but it all sounds like the same autotuned vocals over a high-hat heavy percussive back track! It may be derivative and worthy of criticism, but only in an artistic manner. It is not derivative in a legal manner.
If you work in copyright law, why ask random internet-commenters? The allegations and arguments are the legal filing [0], such as:
> [4] Defendants’ GenAI tools can generate output that recites Times content verbatim [...]
> [5] Defendants also [...] generate responses that contain verbatim excerpts and detailed summaries of Times articles that are significantly longer and more detailed than those returned by traditional search engines. By providing Times content without The Times’s permission or authorization, Defendants’ tools undermine and damage The Times’s relationship with its readers and deprive The Times of subscription, licensing, advertising, and affiliate revenue.
> [6] [T]here is nothing “transformative” about using The Times’s content without payment to create products that substitute for The Times and steal audiences away from it. Because the outputs of Defendants’ GenAI models compete with and closely mimic the inputs used to train them, copying Times works for that purpose is not fair use.
[0] 1:23-cv-11195 - https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec20...
[5] Excerpts and Summaries are fair use, or should be.
[6] Is there a market for a chatbot that simply repeats a single news source? Its not really the purpose of the product.
Imagine that Acme Music hires large sweatshops of workers, who are tasked with listening to trending (copyrighted) songs and then go through grueling practice to sing and play near-perfect performances of those songs, delivered to any customer who asks for one.
Is just plain wrong on how I would interpret the statement.
But I have given you the benefit of the doubt and checked to see if you meant they were literally just re-recording already existing pop songs.
If they were just pumping out artistically derivative drivel of trending styles, but not derivative in terms of copyright law, that’s perfectly legal.
I’ve been asking for clarity, not your opinion on the law.
And even when it comes to copyright law, a good lawyer with enough of their client’s money can make any sort of argument and find a ruling in their favor, but these legal discussions tend to be spoken of in absolutes by random internet commentators!
Perhaps because they're tools, not people.
So why should an artificial brain be treated differently, especially when it's been shown how you can prompt one in a way to almost regurgitate the exact copyright info (similar to how you can prompt a human brain)?
Don't listen to what they say, watch what they do.
If it truely were impossible, then it sounds like they've just admitted that they should have licensed the content rather than using it without permission.
AI's gonna get trained on it no matter what; nobody wants a product that isn't. By the time OpenAI takes the bullet we will have killed nothing.
If you can convince adverts that a scaped hit counts as an organic view, I can concede on your point of "educating yourself on copyrighted material is okay". But given NYT here, that seems to already have been a lost point.
>AI's gonna get trained on it no matter what; nobody wants a product that isn't.
That's fine. Just license it like everyone else that wants to use copyrighted material for commercial use. Why is that so hard? It's not like Microsoft can't bankroll it.
OpenAI's quote, from which the headline is derived, briefly mentions public domain content:
> > Because copyright today covers virtually every sort of human expression including blog posts, photographs, forum posts, scraps of software code, and government documents – it would be impossible to train today’s leading AI models without using copyrighted materials. Limiting training data to public domain books and drawings created more than a century ago might yield an interesting experiment, but would not provide AI systems that meet the needs of today’s citizens.
I do think they're right in that you're not going to get remotely the same performance as modern LLMs if you train with solely public domain content. Licensing a sufficient amount of content could be feasible for OpenAI/Meta, but hopefully we get something that still works for smaller companies, FOSS groups, etc.
Even being in favor of massive copyright reform, I don't think I can get behind that argument.
My intuition is that, if you publish something online and let me read it, I can use that knowledge. I can NOT flat out duplicate it for others, but I can still use the knowledge. Training OpenAI is akin to me using something I read. (Hell, most of what I know I obtained this way.)
To me, the only issue is if the model is reproducing text verbatim en masse.
I think my intuition above is basically where my intuition lands for search engines too. The can use datas found on the web, and can show me snippets, but I'd draw a line at them showing me all/most of an article.
Think of my intuition as an all you can eat buffet. In spirit, you can only eat so much as one person and you may bring home some scraps of food you already touched. In theory (if we ignore fine print), you can very much eat out half the buffet, and then take the other half to go. But this latter "approach" breaks down the model of a buffet. Tragedy of the commons. Everyone is a glutton, the buffet can't keep up with making food, business shuts down, no more food for anyone.
The internet here is a buffet. Just because you have access to terabytes, petabytes or more of information without paying doesn't mean you can download all those terabytes of info. Everyone starts doing that and the model of the commons that are ad-supported servers breaks down. Sites start to close up behind more anti-copy protection or literal paywalls to make things not-so-freely accessible. And that ends the free internet as we know it. We are indeed starting to see pieces of this happen in real time.
>To me, the only issue is if the model is reproducing text verbatim en masse.
That's the other issue as well that was discovered with that exploit. You can reguritate the source with enough prompting. This seems to imply that some server out there is in fact just storing hoards of source material on itself. If that server is commercialized and has TBs of copyright material, that seems to go past the fair use argument of "limited use".
Sites have been pushing towards "to continue reading, download our app" or "you've reached your free article limit, subscribe for just $X" for a long time, because it's profitable. I don't think there's sense in retroactively placing the blame for this behavior on crawlers - it's a tactic designed for humans, to the extent that it's often implemented in ways entirely ineffective against bots such as putting an element in front of the content, using cookies to track viewed article count, or even intentionally letting bot user-agents through the paywall. Sites are capable of restricting GPTBot/CCBot/etc., or putting a rate limit on crawling, without impacting human readers.
> You can reguritate the source with enough prompting. This seems to imply that some server out there is in fact just storing hoards of source material on itself. If that server is commercialized and has TBs of copyright material, that seems to go past the fair use argument of "limited use".
The model isn't retrieving content from some storage server during inference if that's what you're thinking, but consider Google Books which does actually do that - storing a huge number of in-copyright books internally on Google's servers and retrieving snippets when searched. As found in Authors Guild v. Google, this is Fair Use - in particular:
> > What matters in such cases is not so much "the amount and substantiality of the portion used" in making a copy, but rather the amount and substantiality of what is thereby made accessible to a public for which it may serve as a competing substitute.
And websites have been getting paid for a long time by selling user data (which again is only useful for human data, not robot data). I don't think it's a coincidence that the paywalling of more and more sites came as GDPR and other data privacy regulations started to pop up. Crawlers are one of those smaller symptoms that gets knocked out along the way as well. Can't make money one way, they switch to another (even before taking the pandemic and interests rates into account).
>Sites are capable of restricting GPTBot/CCBot/etc., or putting a rate limit on crawling, without impacting human readers.
all those captchas I'm being hit with these days tells me otherwise. Maybe it is possible, but perhaps too expensive. Or the capthas themselves are great for training data. I don't know the real reason, but the internet has definitely become most hostile towards my human eyes in an attempt to curb bots. And the worst part is that it doesn't even entirely block out the bots.
>The model isn't retrieving content from some storage server during inference if that's what you're thinking
I don't admit to know the full details, but some exploits discovered lately seem to suggest that this isn't just "reading a page" and then moving on: https://arxiv.org/pdf/2301.13188.pdf (PDF warning). Apparently somewhere in the memory there is those raw assets being stored for "reference".
Again, I don't know enough to debate on it, but I've seen enough pieces to be scrutinous of how the process currently works. Scraping is a gray area, but storing copyright data in your commercial product is a much harder issue to argue. Youtube screws its creators over backwards to prevent litigation over a very similar concept.
Data protection regulations being the cause of paywalls seems a different theory than what you were previously suggesting.
> all those captchas I'm being hit with these days tells me otherwise. Maybe it is possible, but perhaps too expensive.
Captchas are generally to prevent spambots. CCBot/GPTBot crawlers identify themselves by their user-agent (so the server can just choose not to serve the content) and also respect robots.txt.
> https://arxiv.org/pdf/2301.13188.pdf (PDF warning). Apparently somewhere in the memory there is those raw assets being stored for "reference".
Stable Diffusion is ~4GB and trained on 5 billion images - for the most part it's not plausible for it to have stored the "raw assets". There is a real issue of overfitting on a subset of content that the model saw many times during training such that it was able to memorize details, but the model isn't retrieving the source material from some server to reference (that's known as retrieval augmented generation, which isn't used by default SD/GPT).
Would also keep in mind that for papers like this, where they're generating hundreds of millions of images and digging into CLIP embeddings to construct prompts that intentionally recreate images, that there's some degree of similarity you'd expect just by the conditioning given, the large number of comparisons being automatically made, and coincidence - even had the model not seen the image in question.
> Youtube screws its creators over backwards to prevent litigation over a very similar concept.
Youtube makes the video available to the public. The Google Books analogy properly preserves the aspect that full works are stored but only small portions made available.
Sure. I guess we'll see how that holds up in the lawsuit over the coming years.
> for the most part it's not plausible for it to have stored the "raw assets"
I wasn't talking as literal pngs on a server. I'm sure a company backed by Microsoft has much more clever ways to store such data. My main point was that it seems possible to reproduce the asset so enough of it is kept to perform such operations.
>Would also keep in mind that for papers like this, where they're generating hundreds of millions of images and digging into CLIP embeddings to construct prompts that intentionally recreate images, that there's some degree of similarity you'd expect just by the conditioning given, the large number of comparisons being automatically made, and coincidence - even had the model not seen the image in question.
coincidence, perhaps. I can't say and tracing isn't technically illegal either (although extremely faux pas for artists doing anything more than learning how to draw). But that is definitely where the metaphor of human motor skills and machine reproduction breaks down. Humans can more easily be cast off as coincidence based on context (which can be easy or hard to deduce. E. G. Did artist B follow artist A on Twitter? How long were they active? Are they known to have a similar style?)
bots' context is closed up as company IP. That alone will give it a lot more scrutiny. Scrutiny that can potentiallybe subpeona's, so again: TBD.
>Youtube makes the video available to the public. The Google Books analogy properly preserves the aspect that full works are stored but only small portions made available.
I don't really know why you're so fixated on the Google books metaphor. Google books is also an actual bookstore as well as some place for users to upload their library. I can upload pirated content for private use becsuse Google isn't scanning all its customers 24/7. It's technically not legal but also hard to enforce on my private library of 1 that I share with no one.
I'd run into more issues if I tried to sell it and someone caught me. You can DMCA a book from Google just as easily as a YouTube video. There's simply less money in the game to DMCA troll literature (and for many books of interest to HN, the primary customer isn't individuals but institutions and education).
It's an example we have a court ruling for and is a more literal case of internally storing a huge number of copyrighted works, without permission of the copyright holders, then allowing users to query for snippets.
> Google books is also an actual bookstore as well as some place for users to upload their library. I can upload pirated content for private use becsuse Google isn't scanning all its customers 24/7.
Google Books is a search tool for millions of books that Google have scanned and OCR'd. As far as I'm aware it doesn't itself sell books and I don't believe there's a way for regular users to upload content.
Are you thinking of books on Google Play?
> It's technically not legal [...] You can DMCA a book from Google just as easily as a YouTube video
Google Books has been found to be legal due to Fair Use. You could send a DMCA takedown notice, but they wouldn't have to comply with it.
I think that's a subtly different angle since it doesn't make any appeal to scale/cost. Sort of like the difference between:
1: "Everybody should be free to use Article X in manner Y without negotiation or payment, as a matter of principle."
2: "When I choose to use millions of Article X'es in manner Y, it just becomes expensive and impractical to negotiate/pay, therefore my situation is qualitatively different and exempt."
It's true but still breaking the law.
But I guess the sad fact of it all is that I'm sure they 100% anticipated lawsuits, and expected that fighting them is still cheaper than trying to license all the content properly. I doubt anyone is surprised by this lawsuit. Just disappointed that they will throw money at anyone other than the actual artists and writers.
If they can't manage to strike out deals with content creators despite having M$' coffers at their disposal, then they should cease to exist. Let's also nuke M$ from orbit and split them up into a trillion little pieces while we're at it
Which is why any sane country would extend fair use to all of these things no questions asked, short of 1:1 replication.
Yankistan however is not a sane country and I am expecting the outcome to these legal tests to be absolutely bonkers.
>But they make money
Yes, so does the guy who takes 20 public domain stories, publishes it as a collection on google books and sells it to me for 1 dollar. People act as if their models aren't hugely backed by human labor, labor that is hugely transformative and commands a decent reward.
if that was all they wanted to scrape, I wouldn't mind that. Reddit et al. would have an issue, but they don't exactly compensate commenters to begin with either.
But given that visual works are being copied and that they are being sued by NYT (who hired writers for over a century to produce what is now being scraped for free), it's clear that that wouldn't be enough.
>so does the guy who takes 20 public domain stories, publishes it as a collection on google books and sells it to me for 1 dollar.
curation is a lucrative business in the Information Era, so if there's no licensing issue I don't see the issue. I've spent days, weeks researching various problems and I'd happily pay a dollar for someone to collect a dozen good sources into one place to save me that time.
If all of OpenAI's content was public domain, there would be no lawsuit.
OpenAI's wholesale abuse of the internet has made the internet no longer an acceptable place for me to distribute data and writing to the general public. I have no sympathy whatsoever for them.
For example, I make an open source Firefox web extension for filtering internet content with my own classifier. That literally would not be able to exist without being able to be trained on web content, much of which is copyrighted. Requiring that I somehow either a) use only attributed data or b) detect and not use copyrighted content when trying to build something representative of my source distribution (e.g. the web) sounds like a recipe for a poor outcome. Now maybe my addon isn't your cup of tea - but what if you found out that the next generation of uBlock Origin etc. could not be as effective because of legislation because it wanted to use an AI model? Legislating too heavily around this area will, I believe, have a tremendous chilling effect for small businesses and open source folks trying to innovate in AI.
I've also worked commercially in the creation of two closed source machine learning models, but the domains were restricted enough that web content was not a particularly helpful input. One did all right, and one did not. Seeing bets succeed and fail gives me appreciation for the long-term and uncertain bets that OpenAI has been making for ages finally coming to fruition. I think without businesses being willing to make those bets the GPU-hours would have been hard to pay for.
I've wondered if potentially a different way out of this is not restricting the use of copyrighted material in the training process itself, but rather to instead only consider the created final works. Of course there are thorny problems there, too, but I don't see that having the same chilling effect on research and probably a lesser effect on business as well. One thing I think is clear though: we've reached a tipping point in the US similar to 1998 when the DMCA was legislated where the technology is forcing us to think carefully about what copyright means.
So I have question for those on HN who have meaningfully worked in the creation of not just AI-generated content, but in the creation of some AI model that others use freely or commercially: what seem like promising paths forward here? Or to those working in copyright law (like @williamcotton): how do you see the status quo and potential paths forward?
Any human who wants to access copyrighted works is by law required to honour the copyright - whether that means purchasing of licenses, timed rental access or immediate cease and desist of usage. Why should AI (and the billion-dollar-backed companies building them) get different treatment?
Or to turn it the other way: if the billion-dollar-backed companies building AI models do get a free pass then surely humans should too?
sure, because there's an incentive to attract humans to view such content. It's advertising. Google images isn't built out of goodwill, but is now a target to optimize for to maximize human traffic to get human eyeballs to view ads (or paywalls) for humans consume more products. Having a bot come in ruins that, and the literal billions thrown at adtech to try to verify organic traffic shows that these bots are not desirable metrics for those who pay for ad space.
That perspective from a business lens shows the difference between a human viewing copyright material and a bot. Humans are monetization targets, bots are not. Humans can advertise for you to other humans, bots can't (well... not yet. But do we really want to be talking about ChatGPT ads this early in the LLM era?)
Maybe this is how the news and publishing industries get repaired.
Can't cut into those potential margins after all.
In that case there would be some separate DB used to identify copyright material that it cannot use when generating art. Hopefully with drastic sanctions if it arises that it is in the LLM.
Alternatively, we can go back to literal Youtube and rely on DMCA yet again. I don't think anyone desires that.
I really hope my child (now 8) can live an adulthood free of the ridiculous notion that knowledge can be owned.
You may not be able to own knowledge, but people sure don't want to take the time to accrue such knowledge to learn how to draw themselves. Easier to throw scraps at those who did.
Just send the BSA to visit them. Oh, wait, they _are_ the BSA. /s
How time flies.
They arnt open sourcing their models and they are actually monetizing it. As for AI safety.....its so grey aread. Uncensored models have a place in this community, and trying to suppress them or claim their is no use for them is just insane to me.
What one person views as a moral or ethical response another might view as wrong. This is all culture and also based on your country's views which help form individual perspectives.
I think a big issue with AI other than the hardware barriers is that we have no clue if the amazing developments in the field will keep up with the same pace or stagnate. We can't accurately tell when we have hit the theoretical limit, or when we have true AGI.