Hacker News Reader

Comment by yellowapple | Hacker News Reader

Makes sense to me. AI models are trained on (as large of a subset as possible of) the sum of human knowledge, so their outputs should belong to humanity as a whole.

Really so should all creative works, on the same basis of all creative expression being the product of the society and civilization which fundamentally and inescapably influenced the creator — and they would belong to humanity as a whole, if it wasn't for intellectual property systems demanding the removal of ideas from the commons.

View on HN

>Makes sense to me. AI models are trained on (as large of a subset as possible of) the sum of human knowledge, so their outputs should belong to humanity as a whole.

No it doesn't, because that would mean any sort of secondary source shouldn't be eligible for copyright either, eg. encyclopedias, which are basically rehashing "the sum of human knowledge".

Interesting POV. Remind me - did those encyclopedias pay experts to write their contents, or did they just hoover up every bit of text they could find - regardless of owner - and toss it into a blender?

Sure, I'll remind you. The most popular encyclopedia of our time, Wikipedia, doesn't pay most of its contributors (I think they do have some administrative staff or something?). Nor do they pay journalists or authors for the articles and books that they cite.

Interesting, interesting. And this wikipedia - it gets its contents by hoovering up the web, then? Or do you maybe want to put on your thinking hat and consider the difference between voluntary contributions and theft?

Yes, Wikipedia gets its content largely by hovering up the web, without the consent of the authors. For example you can cite a New York Times article in Wikipedia, without getting the consent of the NYT author. Wikipedia also "hoovers up" (as you put it) offline sources like books. Again without consent!

Indeed, the need to get consent from the author before reading a published work Isn't A Thing in general, outside some very specific contractural scenarios.

Why are you claiming that citations are the same thing as theft without citation? There’s a mile of difference between citing a work and taking it, rewording it, and not crediting or compensating the original author.

> There’s a mile of difference between citing a work and taking it, rewording it, and not crediting or compensating the original author.

So what's wikipedia doing vs what LLMs do? So far as I can tell the only difference is in citations, but:

1. LLMs can be made to cite, eg. if you use google search's AI mode it'll happily provide citations. I doubt that would placate the AI haters though.

2. Outside of academia no one really cares about citations. There's no legal requirement to cite, nor do I think all the people complaining about AI "stealing" other peoples' work are going to be magically placated by the addition of a few citations. Moreover it's unclear whether the concept of citations makes sense in many contexts. If you ask a human programmer how to write fizzbuzz, they'll likely blurt out a solution without providing citations, much like an AI would. Same for most questions people are asking AI about, eg. "gimme a cake recipe", do you really need a citation back to some 18th century cook book?

> So what's wikipedia doing vs what LLMs do?

For Wikipedia, people look up sources to create some original work. They do it while citing the original work but that's not the major point. That's clearly all lawful, you are allowed to do that.

For LLMs, people (or let a software do it, doesn't really matter) copy original work and use it for commercial purpose. If that's allowed fully depends on the license of the original work, but I didn't think anybody really beliefs that OpenAI and others check the license for every single original work they ever used. So it's basically the biggest copyright infringement ever.

Everything else is just framing coming from big companies.

So, the problem already starts while training the model.

Regarding its output, if it happens to output work that falls under copyright, the LLM company must make sure that it obeys the license connected to it (i.e. citing or not relaying the result to the user). Obviously, nobody does that and it's also not generally possible to do that anyway. So that would be second biggest copyright infringement ever that only works because it's hard to track when such an infringement happens.

So, if asked "could we use your work for our commercial software that might output something that would be still protected by your copyright, but nobody will be able if or when it happens and we won't check and won't tell the users" nobody would have given consent. So they went "duck it, we are talking about billions of dollars and AI is great etcetc., so let's just do it anyway"

>For Wikipedia, people look up sources to create some original work. They do it while citing the original work but that's not the major point. That's clearly all lawful, you are allowed to do that.

>For LLMs, people (or let a software do it, doesn't really matter) copy original work and use it for commercial purpose. If that's allowed fully depends on the license of the original work, but I didn't think anybody really beliefs that OpenAI and others check the license for every single original work they ever used. So it's basically the biggest copyright infringement ever.

So what makes wikipedia (and other encyclopedias) legal but chatgpt not legal? By your own admission citation isn't "the major point". Wikipedia might get a pass because it's a non-profit, but every other encyclopedias operate on the same model.

I've explained this many times. LLMs don't "copy original work." ChatGPT doesn't contain copies of books inside it, any more than your brain is a copy of the various books you have read. The LLM model isn't physically big enough for that, it's like saying you fit 1000 gallons of water in a 1 gallon milk bottle. Can't be done. The model may be able to quote small snippets of works, just like you might remember various quotes from Shakespeare or someone. (That's not infringing either, by the way)

Wikipedia, and LLMs, can refuse to cite sources and still not infringe copyright. Citation simply isn't relevant to copyright. Not citing a work that you read previously is not a copyright infringement. Wikipedia or OpenAI being non-profit, or for-profit businesses, has nothing to do with copyright. Consent has nothing to do with copyright. Copyrighting something doesn't mean that you can require everyone who reads it to get your consent. You can require everyone who distributes it to get your consent, but once it's been distributed to someone, they can read it freely.

Hope this helps!

The whole job of a Wikipedia author is to take source material and reword and summarize it into an article. They're essentially acting as human LLMs. Wikipedia even has very specific rules against "original research." You are supposed to be acting as a summarizer, not a researcher or author. Wikipedia also does not compensate the original authors of the source material.

Really the only difference between the Wikipedia author and the LLM is that the Wikipedia author will more frequently be asked to provide citations. But the LLM can also provide citations if asked. In neither case are the authors of what is being summarized compensated or asked for permission. In neither case is it theft.

I don't see how "experts" are relevant under OP's framework unless they're did the primary research themselves. Otherwise they're just regurgitating someone else's research. The "oh there's humans involved so that gets pass" excuse doesn't work either, because humans were also involved in training the AI.

Please reread the comment you are replying to. I will wait.

—-

Now that you have reread the initial comment, do you think that “experts” was the important part? Or do you think maybe it was the compensation for their work that matters?

>Please reread the comment you are replying to. I will wait.

If you're going to post thinly veiled implications that I didn't read your comment, you should be pretty damn sure that you make it look like you read my comment, which it doesn't seem like you did. The second of my comment said:

>The "oh there's humans involved so that gets pass" excuse doesn't work either, because humans were also involved in training the AI.

If you did read it, you sure did a poor job at rebutting it, leaving it unaddressed and preferring to waste words on writing snarky remarks instead.

> If you're going to post thinly veiled implications that I didn't read your comment

It was a statement, not an implication.

You still haven't replied to the point in the original comment about compensating the people who do this work, so I think it's quite obvious that you haven't read it.

>You still haven't replied to the point in the original comment about compensating the people who do this work, so I think it's quite obvious that you haven't read it.

Issac newton discovers the theory of gravity. He advanced the sum of human knowledge, so fair enough, he should get compensated.

Alice rehashes that and puts it into her encyclopedia, allowing others to learn the theory of gravity.

Bob writes an algorithm for training a chatbot that can produce responses rehashing the theory of gravity, also allowing others to learn the theory of gravity.

Why should Alice be compensated but not Bob? Neither discovered the theory of gravity, so it's not like by funding Alice we're helping discover quantum physics or whatever. It's also not obvious that Alice's work is more valuable. A chatbot interface is often better at teaching someone than a rehashed overview. Of course, you can try to fix this by declaring that human work is valuable and an AI model isn't, by fiat, but that's just a cope and a far cry from the original principle of "trained on [...] human knowledge, so their outputs should belong to humanity"

None of this matters for applying the law, because the law just says only human created works are eligible for copyright protection, but that's not the argument OP was trying to invoke.

Finally none of this actually matters because OP just bites the bullet and says that secondary sources shouldn't be eligible for copyright, period.

Fun fan fiction, bro. Stealing content to train LLMs is still stealing.

Correct, and if it wasn't already obvious by now I believe that to be an unambiguously good thing.

Like, now with the advent of AI, or even before? You might not have much love for encyclopedias, which were mostly replaced by wikipedia, but the "secondary sources don't get copyright protection" would also cover programming books, which roughly speaking are docs rewritten to a cohesive narrative.

> Like, now with the advent of AI, or even before?

My disdain for intellectual property predates the existence of LLMs by at least a decade.

It's either/or. Either you reward creators and inventors to keep creating and inventing and keep poverty at bay, or you give everyone enough to avoid poverty whether they work or not and reward c+i in some other way.

What we have now is neither - owners are hugely over-rewarded for owning things and extracting passive value from everyone else, creators and inventors are kinda sorta rewarded sometimes if they're lucky and very much not if they're not. Just like other workers.

"The commons" is not a thing in this model, except in a few small niches.

>What we have now is neither - owners are hugely over-rewarded for owning things and extracting passive value from everyone else, creators and inventors are kinda sorta rewarded sometimes if they're lucky and very much not if they're not. Just like other workers.

Creators and inventors are rewarded but obviously they cannot consume the whole pie. The people who invest in creative pursuits eat a lot of losses. People only seem to notice profitable successes, and forget that failures need to be paid for as well.

The same logic also applies to workers. The fact that your labor costs money is a guarantee, and it might not make money at all. We can think of a few examples where the work is directly delivered to consumers with zero marginal overhead, but most work DOES have overhead and liabilities, no matter how simple.

That's indeed one of many reasons why I'm staunchly pro-UBI.

I don't really understand this, I was trained on all the knowledge I was capable of ingesting, but my outputs are largely mine unless there is too much similarity to a copyrighted work. I don't understand why AI would be different. When I do work for hire, my employer owns it, roughly equivalent to me paying Open AI/Anthropic for output.

It's not, barring any secondary issues like proof the training data was acquired illegally.

When there is a copyright dispute, it's always a subjective test on the judge/jury's part as to how similar it is in appearance, purpose, etc. if it's not deemed fair use.

I’m all for this, if we also set up a system to let society pay for artists. And I mean in full: cover for their living expenses; their rent; internet access; materials; everything.

Or how do you suppose art gets created for humanity as a whole to enjoy?

That, too, is something for which I've been advocating for quite some time now.

>And I mean in full: cover for their living expenses; their rent; internet access; materials; everything.

Who decides which artists get subsidized? How would larger projects work? Seeing how often open source volunteer projects implode due to various community drama, this sort of system would basically preclude any sort of big production.

> Who decides which artists get subsidized?

The easy answer is to just pay literally everyone. This is called a “universal basic income”, and has been repeatedly demonstrated to be a good idea for many reasons besides decoupling creative expression from the need to put food on the table.

Really gotta force the starving artist motif or what's the point?

That we all stand on the shoulders of giants? There is no private innovation that isnt built on mountains on publicly shared knowledge and innovation, and usually public funds as well. The idea that private enterprise leads innovation is propaganda, i.e publicly funded academic study paved the way for LLMs, corps just commodotise ideas to make them viable under capitalism.

>The idea that private enterprise leads innovation is propaganda, i.e publicly funded academic study paved the way for LLMs, corps just commodotise ideas to make them viable under capitalism.

So if you founded some wildly successful unicorn, I (or the government) can come over and say "nice startup, too bad it runs off of the internet (based off ARPANET), so your startup belongs to the the state now"?

Got a whole wagon of straw to sell you, you're going to need it

The government can already do that if it so chooses, given that the government is the singular reason why any corporation is able to exist as an independent legal entity in the first place (as opposed to a bunch of individual persons collaborating informally).

Reply on news.ycombinator.com