back

by ninjahawk1·24d ago·view on hn ↗
The U.S. does NOT have copyright protections on the data that AI models were and are trained on, it was mass theft. Now all of a sudden, the Chinese are stealing from the thieves and we’re supposed to care?

No one cares, we just want cheaper AI models that are equally as powerful. They’re all thieves regardless.

12 comments
Ok, but can you please follow the HN guidelines and focus on specific, thoughtful discussion, not generic/indignant tangents? The first is what we're trying for, and the second basically destroys it.

"Eschew flamebait. Avoid generic tangents." - https://news.ycombinator.com/newsguidelines.html

Ultimately this is more a problem with the upvoting system than the comments themselves, since generic/indignant comments routinely attract lots of upvotes, and then sit on top of the thread, smothering more interesting discussion. But in terms of moderation we can only reply to commenters.

Its even dumber when you consider AI model output was not considered copyrighted or eligible for copyright to begin with, as far as I know. How exactly would this even be "theft"?
Plus on top of that it's been ruled that training on copyrighted material is allowed because it's transformative. As a result even if the model output is copyrighted training on it would be as allowed as training on books.
I can't imagine Moonshot have found a way to completely bypass the subscription or api billing either. So those model outputs have been paid for! No theft of any kind!

At most they'd be getting more out of distributed use of subsidized plans and API than the ToS would prefer.

It’s theft because a lot of rich people have money on the line.
It’s not they, but it does break TOS.
steal from the rich -> theft

steal from the poor -> capitalism

Training on copyright was ruled to be transformative.

Anthropic was just fined for not paying for the pirated books they copied into a training database. Same as anyone else who copied pirated IP onto their hard drive.

However, training an AI on copyright has been ruled to be sufficiently transformative and not a violation of IP laws. The same way you can make a gameplay clone of Call of Duty without any issue.

So they got fined for… stealing?

Also, if I clone Call of Duty, and use their skins, make a similar soundtrack, call my maps the same, then there will definitely be an issue.

I support the use of AI and all, but to hide behind transformative use of copyrighted intellectual property is a discredit to the colossal amount of human work and knowledge that these companies pirated and had their models trained on.

By that same argument, distilling a model to train your own is also transformative. We can’t have it both ways. Either both training on copyright and distillation are transformative, or both are illegal. Also, it is not a matter of settled law that training on copyright is transformative. That was a single case, and others are still pending AFAIK.
>"Anthropic was just fined for not paying for the pirated books they copied into a training database."

Normal person would be sitting in jail for doing same thing. It is a shame they call it justice.

> Same as anyone else who copied pirated IP onto their hard drive

Not really the same. Private citizens have received fines MUCH higher per-work when downloading for just their private consumption.

Ruled by whom? The International Court of the Northern District of California?
What if use an LLM and attach CoD screenshots, or assets?
You're trying to kidnap what I've rightfully stolen!

https://www.dailymotion.com/video/x4cxcsa

Are public libraries thieves?

Is Anna's Library thieves?

Is Library Genesis thieves?

Is Archive.org thieves?

Is PirateBay thieves?

They all look like public libraries to me. And better access to all human knowledge is a net positive for everyone.

Public library buy the works they distribute. They also pay for the usage of these works. Copying of said works is still illegal. You should read up on how libraries actually work.
That’s fine too, in that case Open Weight models made by China are doing humanity a favour
Someone going to public library, copying every book out of it and then selling them as their own are thieves.
Copyright infringement, maybe. Theft? No. When something is stolen from you, you don't have that thing anymore.
The answer to widespread theft is not to simply allow the guy who did the theft to get away with it because his implementation was cheaper.

It's to punish theft.

But distillation is not theft. They are paying for the usage of the API. It is also not copyright infringement, as AI companies do not hold the copyright for the output of their models.
Exactly! While the people who spent more money on their theft are rewarded with more business to counteract their losses! Justice is finally being wrought on those who have wronged society with their wicked ways!

(/s if it wasn't obvious)

Being serious, it is far more likely, especially given the events of the past 18 momths, that Trump is just doing autocrat things because he can, and ultimately doesn't really care what is and is not good for American business as long as he continues to grow his wealth through corrupt behavior.

not when by your own rules it's legally ruled "not theft".
> The U.S. does NOT have copyright protections on the data that AI models were and are trained on, it was mass theft.

Absolutely false. Courts have decided that it is fair use [1]. That makes sense. It should not be used to justify Chinese theft.

https://www.reuters.com/world/us-judge-approves-anthropics-1...

Why is it not fair use for the Chinese models if all they are doing is distilling these fair use models? According to the courts, AI model output is not copyrighted or eligible for copyright.
Correct. This seems to be poorly understood.

You can use an original work to write an encyclopedia (e.g. Wikipedia). That is fair use.

You cannot copy an original work and distribute it.

LLMs are an encyclopedia.

This is a silly argument. Like yes, obviously it’s of less U.S. concern if one American is robbing another versus a foreign interest. We can deal with one whenever while the other creates constraints.

The better arguments are the ones made in the article. Open weights increase competition and thus AI availability in the U.S. economy.

This is not robbing. Undesired distillation is a ToS violation, at best.

There is not even jurisdiction over Chinese companies, because they don't sell their LLM subscribtions, unlike US companies, and those comitted much more severe violations in getting their training material than a ToS violation or two (Anthropic already found guilty).

edit: I don't see this as legal argument against the ban, more like a justification for why basically no moral person is gonna side with OpenAI and Anthropic on this. US gov can try and ban as many weights as they want, I expect that to be similarly effective as banning numbers was in the past (i.e. not).

It's not like those American companies were only stealing form Americans. Maybe you could argue about utility, but morally I don't see any ground to stand on

And the utility argument is also very shaky. The proposed punishment for foreign interests stealing "US data" is that other Americans now aren't allowed to benefit from that, while the rest of the world can

Obviously stealing is bad, doesn't matter who you steal from. In this case, its a simple case of no honour amongst thieves.
Whether or not it’s robbery is irrelevant. It’s the proposed remedy that is the problem
Blatantly lack of principles.
The era of so-called "intellectual property" is over.

You can accept it or not. But it's going to be a lot easier if you do.

So how come Disney still gets upset when I provide copies of their movies on my website as "training set for biological neural networks"?

The problem isn't that IP is dying - in most industries it died ages ago. The problem is that it is now explicitly reserved for the rich.

One’s like taking something from your brother or sister and the other is like taking from your neighbor with whom you have a bad history…
I do care.

It's an unfortunate truth that we're exiting a global peace era and entering one of (hopefully just...) implied conflict. Our biggest geopolitical rival that's been threatening a hot war also distilling the things we build is not a good thing.

Trump is a huge self-own by our electorate sadly, but being concerned about a conflict between China and the US is a very bi-partisan concern for good reason.

Luckily for the peace side, the USA's inability to achieve their mission objectives in a war against an adversary that was perceived as being substantially weaker means they are very much less likely to engage in one with a "near peer".

The take away from modern conflicts is that massive military size disparities don't translate into successful invasions anymore.

In other words, it's unlikely USA will pursue a conflict with China, and it's unlikely China will pursue a conflict with Taiwan.

Might wanna reflect on who’s threatening hot war lately.
Our biggest geopolitical rival is also the only thing standing in the way of a potential duopoly on LLMs using closed weight models. I would not say I’m a fan of the PRC by any means, but I also do not want to see closed weight models controlled by two big companies dominating what I can use. Especially since said companies arbitrarily restrict model usage and can ban users for no reason.
> It's an unfortunate truth that we're exiting a global peace era

You’re a tiny bit late on that one

I keep hearing about this "we" as if it's still a coherent nation state and not just an unaccountable ruling class from both parties where the most severe consequence is failing upwards into a lobbyist provided position. I literally have no moral preference when it comes to the US or the CCP nor do I care if the entire nation state fails, let alone loses to a geopolitical rival.
I honestly wonder what world folks like you live in. The US is currently at war, repeatedly threatening several other places and engaged in several costly trade wars. From my Canadian perspective, you are a danger and a menace. China is a commodity supplier. Do the war crimes your country is currently committing not register for you at all? The daily threats, insults, and economic attacks on your 'allies', does that register?

China supports Russia in its war formally, but practically it sells Ukraine whatever it needs, because a weak vassal-like Russia is very much in their interests. The US is far, far more outspoken and active supporter of Russia in this war.

Finally, there will be no hot war with China. They'll likely blockade and invade Taiwan before the end of the decade, but the US will not do anything, partly because it'll be exhausted from it's never ending middle east wars, partly because it'll only take a small bribe to get Trump to back off. He's genuinely fond of dictators like Xi.

I don't think I would characterize this last era as one of global peace.
I’m not convinced that it’s a bad thing. What is your actual argument?