So why are Meta, and say, Sci-Hub treated so differently?
They are not. Meta is being sued, just like Sci-Hub was sued. So, one difference is that the suit involving Meta is still ongoing.
In any case, Meta did not create the dataset. IDK if they even shared it. The researcher who did is also being sued. The dataset has been taken down in response to a copyright complaint. IDK if it is available anywhere anymore. So the dataset was treated just like Sci-Hub. The sharing of the copyrighted material was stopped.
Meta downloading these books for AI training seems fairly straight-forward fair use to me. I don’t see how what Meta did is anything like what Sci-Hub did.
So ISPs are blocking Meta for their breaking of copyright?
Because ISPs block Sci-Hub.
No, one of them is having governments trying to kick off the internet, and the other is allowed to continue doing what they’re doing and the worst they’ll face is a fine. Not even close to the same, completely disproportionate. If they were blocking all Meta LLMs until they had removed all copyrighted material, maybe we could say the same.
ISPs may block sites to prevent unauthorized copying. It’s not a punishment for past wrong-doing. I’m not sure about the details, I think this differs a lot between jurisdictions. But basically, as ISPs they are involved in the unauthorized act of copying. Their servers copy the data to the end user/customer. So, they may be on the hook for infringement themselves if they don’t act.
Again, I am not aware of Meta sharing the copyrighted books in question. So, I don’t know what the legal basis for blocking Meta would be. If ISPs block a site without a legal basis, they are probably on the hook for breach of contract.
IDK on what basis the sharing of Meta’s LLMs could be stopped. If anyone could claim copyright it would be Meta itself and they allow sharing them. (I have doubts if AI models are copyrightable under current US law.)
https://www.nytimes.com/2024/01/08/technology/openai-new-york-times-lawsuit.html
In its lawsuit Wednesday, the Times accused Microsoft and OpenAI of creating a business model based on “mass copyright infringement,” stating that the companies’ AI systems were “used to create multiple reproductions of The Times’s intellectual property for the purpose of creating the GPT models that exploit and, in many cases, retain large portions of the copyrightable expression contained in those works.”
Publishers are concerned that, with the advent of generative AI chatbots, fewer people will click through to news sites, resulting in shrinking traffic and revenues.
The Times included numerous examples in the suit of instances where GPT-4 produced altered versions of material published by the newspaper.
In one example, the filing shows OpenAI’s software producing almost identical text to a Times article about predatory lending practices in New York City’s taxi industry.
But in OpenAI’s version, GPT-4 excludes a critical piece of context about the sum of money the city made selling taxi medallions and collecting taxes on private sales.
In its suit, the Times said Microsoft and OpenAI’s GPT models “directly compete with Times content.”
If the New York Times’ evidence is true (I haven’t seen the evidence, so I can’t comment on veracity), then you can recreate copyrighted works with LLMs, and as such, they’re doing the same thing as the Pirate Bay, distributing copyrighted works without authorization and making money off the venture.
So far, no ISPs are blocking Meta for this.
Meta downloading these books for AI training seems fairly straight-forward fair use to me.
They pirated the books. Is that not legally relevant?
“Straight-forward” may be too strong regarding these books. If they inadvertently picked up unauthorized copies while scraping the web, that would definitely not be a problem. That’s what search engines do.
The question is if it is a problem that the researchers knowingly downloaded these copyrighted texts. Owners don’t seem to go after downloaders. IDK if there is case law establishing that the mere act of downloading copyrighted material is infringement. I don’t think there’s anything to suggest that knowing about the copyright status should make a difference in civil law.
In any case, researchers must be able to share copyrighted material, not just for AI training but also any other purpose that needs it. If this is not fair use, then common crawl may not be fair use either. IDK if there is case law regarding the sharing of copyrighted materials as research material, rather than for their content. But I find it hard to see how it could not be fair use, as the alternative would be extremely destructive. So even if the download would normally be infringement, I doubt that it is in this case.
Eventually, we are only talking about a single copy of each book. So, even if researchers were forced to purchase these books, all of AI training would yield only a few extra sales for each title. The benefit to the owners would be very small in relation to the damage to the public.
