Anyone tempted to double down on this: sure, maybe, someday it’s like The Matrix or whatever. I was 12 when it came out & understood that was a fictional extreme. You do too. And you stumbled into a better analogy than slavery in 1800s.
>harassed
This just in, anonymous forum user SHOCKINGLY HARASSED, PELTED with HIGH-SPEED ideas and arguments, his positions BRUTALLY ATTACKED and PUBLICLY DEFACED.
Post you’re replying to:
Which is what? I’m honestly unsure. Could be: we need to nuke the data centers, or unseat any judge that has allowed this, or somehow move the law from “it’s cool to do matmuls with text as long as you have the right to read it.” Not against any of those but I’m sure I’m Other Team coded to you given the amount of harassment you’ve done in this thread to me and others.
All of this meta-discussion of debating when/how we should debate or complaining how "rude" or "boorish" we are is more off-topic, boring and distracting than any flamewar.
The whole point of this site is to coerce public-facing people (like VC founders) into interacting with anonymous users and debating the ideas themselves. A sort of middle ground between reddit and 4chan. That's what all the "civility" and "kindness" talk is for.
It’s really hard to parse this thread because you and the other gentleman keep telling anyone who engages they aren’t engaging.
You both seem worked up and perceiving others as disagreeing with you wholesale on the very concept that AI companies could be forced to compensate people for training data, and morally injuring you.
Your conduct to a point, but especially their conduct, goes far beyond what I’m used to on HN. I humbly suggest you decouple yourself a bit from them, you really did go too far with the slavery bit, and it was boorish to then make child porn analogy.
All we have is an exquisite, thoughtful, nuanced, analogy of how it is exactly like America enslaving Black people in the 1800s. i.e. a cheap appeal to morality.
Then, it is followed by repeated brow-beating comment to anyone who replied, complaining something wasn’t being engaged with.
What exactly wasn’t being engaged with?
It is still unclear.
Do feel free to share, or apologize even. It’s understandable you went a bit too far because you really do feel it’s the same as slavers in the 1800s in America, what’s not understandable is complaining no one is engaging correctly.
If you distribute child porn, that is a crime. But if you crawl every image on the web and then train a model that can then synthesize child porn, the current legal model apparently has no concept of this and it is treated completely differently.
Generally, I am more interested in how this effects copyright. These AI companies just have free reign to convert copyrighted works into the public domain through the proxy of over-trained AI models. If you release something as GPL, they can strip the license, but the same is not true of closed-source code which isn't trained on.
Basically the exact same thing.
If I had a photographic memory and I used it to replicate parts of GPLed software verbatim while erasing the license, I could not excuse it in court that I simply "learned from" the examples.
Some companies outright bar their employees from reading GPLed code because they see it as too high of a liability. But if a computer does it, then suddenly it is a-ok. Apparently according to the courts too.
If you're going to allow copyright laundering, at least allow it for both humans and computers. It's only fair.
Right, because you would have done more than learning, you would have then gone past learning and used that learning to reproduce the work.
It works exactly the same for a LLM. Training the model on content you have legal access to is fine. Aftwards, somone using that model to produce a replica of that content is engaged in copyright enfringement.
You seem set on conflating the act of learning with the act of reproduction. You are allowed to learn from copyrighted works you have legal access to, you just aren't allowed to duplicate those works.
If someone hires me to write some code, and I give them GPLed code (without telling them it is GPLed), I'm the one who broke the license, not them.
I don't think this is legally true. The law isn't fully settled here, but things seem to be moving towards the LLM user being the holder of the copyright of any work produced by that user prompting the LLM. It seems like this would also place the enfringement onus on the user, not the provider.
> If someone hires me to write some code, and I give them GPLed code (without telling them it is GPLed), I'm the one who broke the license, not them.
If you produce code using a LLM, you (probably) own the copyright. If that code is already GPL'd, you would be the one engaged in enfringement.
LLMs don't "learn" but they _do_ in some cases, faithfully regurgitate what they have been trained on.
Legally, we call that "making a copy."
But don't take my word for it. There are plenty of lawsuits for you to follow on this subject.
"Learning" is an established word for this, happy to stick with "training" if that helps your comprehension.
> LLMs don't "learn" but they _do_ in some cases, faithfully regurgitate what they have been trained on.
> Legally, we call that "making a copy."
Yes, when you use a LLM to make a copy .. that is making a copy.
When you train a LLM... That isn't making a copy, that is training. No copy is created until output is generated that contains a copy.
Only by people attempting to muddy the waters.
> happy to stick with "training" if that helps your comprehension.
And supercilious dickheads (though that is often redundant).
> No copy is created until output is generated that contains a copy.
The copy exists, albeit not in human-discernable form, inside the LLM, else it could not be generated on demand.
Despite you claiming that "It works exactly the same for a LLM," no, it doesn't.
"Learning" was used by the person I responded too. If you had read my comment with any care you would have realized I used the words "training" and "learning" specifically and carefully.
> The copy exists, albeit not in human-discernable form
That doesn't count as a "copy" since it isn't human-discernable. You can't show me the copy all you can do is say "it has to be in there somewhere." It is pretty clear this is a transformative use and so far the courts have agreed
> supercilious dickheads
Very mature.
If you don't like being called out for lack of comprehension, then don't needlessly impose a semantic interjection when you haven't bothered to understand the discussion.
Not in the same sense.
> If you had read my comment with any care you would have realized I used the words "training" and "learning" specifically and carefully.
This is completely belied by "It works exactly the same for a LLM."
> That doesn't count as a "copy" since it isn't human-discernable.
That's not the reason it _might not_ count as a copy (the law is still not settled on this, and all the court cases have lots of caveats in the rulings), but thanks for playing.
> If you don't like being called out for lack of comprehension, then don't needlessly impose a semantic interjection
If you want to not appear mendacious, then don't claim equivalence between human learning and machine training.
> It is pretty clear this is a transformative use and so far the courts have agreed
In weak cases that didn't show exact outputs from the LLM, yes. In any case, "transformative" does not automagically transform into fair use, although it is one considered factor.
> Very mature.
Hilarious, coming from the one who wrote "if it helps your comprehension."
You must be one of those assholes who think it's OK to say mean things if you use the right words.
Bless your heart.
I specifically used the word "training" in the sentence aftwards. "It" clearly refers to the sentence prior which explains that infringement happens when the copy is created, not when the original is memorized/learned/trained.
> If you want to not appear mendacious, then don't claim equivalence between human learning and machine training.
I never claimed that. I already clarified that with my previous comment. Instead of bothering to read and understand you have continued to call names.
> Hilarious, coming from the one who wrote "if it helps your comprehension."
You seemed confused, you still seem confused. If you think this genuine (and slightly snarky) offer to use terms that sidestep your pointless semantic nitpick is "being an asshole"... then you need to get some more real world experience.
If you think my behavior here was truly ban worthy than do it because I don't see anything in the I would change except for engaging at all
> Instead of bothering to read and understand you have continued to call names.
> You seemed confused, you still seem confused
> your pointless semantic nitpick
> you need to get some more real world experience
I wouldn't personally call that being polite, but whatever we call it, it's certainly against HN's rules, and that's what matters.
Edit: This may or may not be helpful (probably not!) but I wonder if you might be experiencing the "objects in the mirror are closer than they appear" phenomenon that shows up pretty often on the internet - that is, we tend to underestimate the provocation in our own comments, and overestimate the provocation in others' comments, which in the end produces quite a skew (https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...).
I know moderation is a tough gig.
If we are saying that the LLM learns things and then made the copy, then the LLM made the crime and should receive the legal punishment and be sent to jail, banning it from society until it is deemed safe to return. It is not like the installed copy is some child spawn from digital DNA and thus the parent continue to roam while the child get sent to jail. If we are to treat it like a living being that learns things, then every copy and every version is part of the same individual and thus the whole individual get sent to jail. No copy is created when installed on a new device.
Right, because then we have to decide at what point our use of AI becomes slavery.
It's also an interesting double standard, wherein if I were to steal OpenAI's models, no AI worshippers would have any issue condemning my action, but when a large company clearly violates the license terms of free software, you give them a pass.
If GPT-5 were "open sourced", I don't think the vast majority of AI users would seriously object.
Which is funny since that's a much clearer case of "learning from" than outright compressing all open source code into a giant pile of weights by learning a low-dimensional probability distribution of token sequences.
Information wants to be free.
That is not nearly the extent of AI training data (e.g. OpenAI training its image models on Studio Ghibli art). But if by "gave their work away for free" you mean "allowed others to make [proprietary] derivative works", then that is in many cases simply not true (e.g. GPL software, or artists who publish work protected by copyright).
I mean, this is an ideological point. It's not based in reason, won't be changed by reason, and is really only a signal to end the engagement with the other party. There's no way to address the point other than agreeing with them, which doesn't make for much of a debate.
> an 1800s plantation owner saying "can you imagine trying to explain to someone 100 years from now we tried to stop slavery because of civil rights"
I understand this is just an analogy, but for others: people who genuinely compare AI training data to slavery will have their opinions discarded immediately.
The idea that they are coming up with all this stuff from scratch is Public Relations bs. Like Arnold Schwarzenegger never taking steroids, only believable if you know nothing about body building.
If an LLM reads a free wikipedia article on Aladdin and adds a genie to it's story, what copyright law do you think has been broken?
To go into details though, under copyright law there's a clause for "fair use" under a "transformative" criteria. This allows things like satire, reaction videos to exist. So long as you don't replicate 1-to-1 in product and purpose IMO it's qualifies as tasteful use.
If a person "trains" on other creatives' works, they can produce output at the rate of one person. This presents a natural ceiling for the potential impact on those creatives' works, both regarding the amount of competing works, and the number of creatives whose works are impacted (since one person can't "train" on the output of all creatives).
That's not the case with AI models. They can be infinitely replicated AND train on the output of all creatives. A comparable situation isn't one human learning from another human, it's millions of humans learning from every human. Only those humans don't even have to get paid, all their payment is funneled upwards.
It's not one artist vs. another artist, it's one artist against an army of infinitely replicable artists.
What is the basis that an LLM should be included as a "creative type"?
LLMs seem to match.
> We have evidence of LLMs reproducing code from github that was never ever released with a license that would permit their use. We know this is illegal.
What is illegal about it? You are allowed to read and learn from publicly available unlicensed code. If you use that learning to produce a copy of those works, that is enfringement.
Meta clearly enganged in copyright enfringement when they torrented books that they hadn't purchased. That is enfringement already before they started training on the data. That doesn't make the training itself enfringement though.
What kind of bullshit argument is this? Really? Works created using illegally obtained copyrighted material are themselves considered to be infringing as well. This is both common sense and law. I cannot believe the mental gymnastics here.
That isn't true.
The copyright to derivative works is owned by the copyright holder of the original work. However uaong illegaly obtained copies to create a fair use transformative work does not taint your copyright of that work.
> Even if not, you agree that they infringed on copyright of something close to all copyrighted works on the internet and this sounds fine to you?
I agree that they violated copyright when they torrented books and scholarly arguments. I don't think that counts at "close to all copyrighted works on the Internet".
> The consequences and fines from that would kill any company if they actually had to face them.
I don't actually agree that copyright that causes no harm should be met with such steep penalties. I didn't agree when it was being done by the RIAA and even though I don't like facebook, I don't like it here either.
>It's a CRYSTAL CLEAR violation of the law
in the court of reddit's public opinion, perhaps.
there is, as far as I can tell, no definite ruling about whether training is a copyright violation.
and even if there was, US law is not global law. China, notably, doesn't give a flying fuck. kill American AI companies and you will hand the market over to China. that is why "everyone just shrugs it off".
what do you picture happening if Western AI companies cease to operate tomorrow and fire all their researchers and engineers?
> There's no way to address the point
That's you quitting the discussion and refusing to engage, not them.
> have their opinions discarded immediately.
You dismiss people who disagree and quit twice in one comment.
I have no interest in the rest of this argument, but I think I take a bit of issue on this particular point. I don't think the law is fully settled on this in any jurisdiction, but certainly not in the United States.
"Reason" is a more nebulous term; I don't think that training data is inherently "theft", any more than inspiration would be even before generative AI. There's probably not an animator alive that wasn't at least partially inspired by the works of Disney, but I don't think that implies that somehow all animations are "stolen" from Disney just because of that fact.
Obviously where you draw the line on this is obviously subjective, and I've gone back and forth, but I find it really annoying that everyone is acting like this is so clear cut. Evil corporations like Disney have been trying to use this logic for decades to try and abuse copyright and outlaw being inspired by anything.
> I don't think that training data is inherently "theft", any more than inspiration would be even before generative AI. There's probably not an animator alive that wasn't at least partially inspired by the works of Disney ...
Sure, but you can reason about it, such as by using analogies.
You cant be serious
i.e. people who are not hackers. Many (most?) hackers have been against the idea of copyright and intellectual property from the beginning. "Information wants to be free." after all.
Must be galling for people to find themselves on the same side as Bill Gates and his Open Letter to Hobbyists in 1976 which was also about "theft of people's labor".