There is no proof so far that openAI has violated any copyright. I understand the public’s emotions but recall that these large language models can read all the public info and make useful generalizations.
Being public doesn’t mean copyright goes away. If you can’t use a 3rd party program even once without permission why would you assume you could use someone’s book etc to train a model without similar permission?
My guess is there’s going to be an exception given, but that’s not yet decided.
Because the law has never delt with the concept of training models like these before and it's not just automatically the same thing as republishing something non-transformative no matter how much some people want to just assert that it is without actually thinking about it. It's a new case, that probably needs new different types of protections and consideration.
Does the law care about the method(whether it is ML or just reading the article from a database) if the model can spit out a verbatim copy of an article?
Researchers have gotten large chunks of verbatim text. Suggesting it’s only infringing if the entire work is reproduced is legally incorrect and therefore irrelevant.
> Suggesting it’s only infringing if the entire work is reproduced is legally incorrect
It may not be conclusive, but it's highly suggestive:
> Amount and substantiality of the portion used in relation to the copyrighted work as a whole: Under this factor, courts look at both the quantity and quality of the copyrighted material that was used. If the use includes a large portion of the copyrighted work, fair use is less likely to be found; if the use employs only a small amount of copyrighted material, fair use is more likely.
I've seen many of these "AI" offerings regurgitate their training data verbatim. If chatgpt were miraculously immune to this, despite taking the same basic approach as everybody else, and without significant efforts to prevent regurgitation, then the onus is on them to prove that.
Algorithms alone don’t qualify as transformative, which is why making an MP3 doesn’t count as a new work.
Copyright is insidious because people tried all sorts of ways to get around it in the past. You can’t for example replace all the words in an article with synonyms and get a completely new work. Thus, asking ChatGPT to summarize an article is problematic because the output is based on the original work irrespective of what’s happening internally.
“At the same time, she also gives the reader the opportunity to see that the government has dramatically changed what its citizens perceive as their history. For example, Montag never knew that firemen used to fight actual fires...”
> Thus, asking ChatGPT to summarize an article is problematic because the output is based on the original work irrespective of what’s happening internally.
You are confusing training a model vs using it. This way a web browser can also be said to use NYT's content.
That’s in support of copyright being insidious not suggesting summaries are the only issue.
The point was people seem to assume only direct copies are problematic but that’s not the case. Training/feature extraction is just as mechanical as making an MP3.
> If you can’t use a 3rd party program even once without permission why would you assume you could use someone’s book etc to train a model without similar permission?
Because you're trying build a silicon brain, and a carbon-based brain is allowed to learn from books and articles without permission (e.g. reading them at the library or online), so why shouldn't a silicon brain also be allowed to learn that way?
In a discussion, one should strive not just to stick to saying things that are true, but to make sure that what they're saying is both true and relevant.
People should obtain their learning material legally. True statement (at least as "true" as any opinionated value judgement can be, I guess). The question is how, though, is that statement relevant here?
Even by the NYT's own telling (the way that OpenAI obtained their articles was from the NYT's website) what OpenAI did was not illegal copyright infringement. The problem is that NYT is under the impression that allowing search engines to access their paywalled stuff and hoping they don't do anything else with it besides putting up conventional, Google-style SERPs makes for copyright infringement if what one of them actually wants it for involves other reasons. It doesn't. There's an adequate legal instrument available for NYT to use if they want to enforce conditions on use: a contract. Do they have a contract that somebody violated? If not, they have no cause to go after anyone. And if they do have a contract that was violated, then that's still not copyright infringement—it's a breach of their contract.
You seem to be confused, giving someone permission to do X is only permission to do X and nothing else.
If I open a store you have implied permission to enter the premises, but not stay inside for 16 hours. Further the second I ask you to leave you no longer have permission to be on the premises. The store doesn’t need to add a lock or put up signs, you’re aware you don’t have permission and that’s it.
Ditch the swipes; I'm not the one confused here. We're talking about copyright infringement, not trespass.
(Your analogy is bad and doesn't hold up. Copyright doesn't grant rightsholders control of the sort required here. It grants them the right to make and distribute copies. It doesn't grant the right to undistribute copies when it turns out they don't like what someone is doing with them.)
A core argument OpenAI is making is transitory copying is allowed as long training is fair use. But a permanent copy stored in a training database isn’t transitory and would itself be copyright infringement, so they don’t do that.
Thus training each version requires permission to download a new copy, which they now lack.
EDIT: The comment I first responded to (only partially reconstructible from quoted parts below) has been edited to something to something almost completely different. Very uncool.
> ChatGPT keeps redownloading works to avoid the issue of keeping permeant copy’s of the training material
I don't know enough about how ChatGPT works to know whether or not that's true, but from an engineering standpoint it certainly sounds wrong because of how insane it would be if true, and I'm not at all convinced that you're right about this given how poorly you understand the other stuff you're trying to argue that I happen to know you're wrong about, so it doesn't seem unwise to conclude that the same is probably true about your claims here. But it doesn't matter, anyway:
> which would be obvious copyright infringement
Wrong. OpenAI keeps asking for copies, NYT keeps giving them out (whether/despite OpenAI having/not having "permission" or not). Not copyright infringement, let alone "obvious copyright infringement".
This is going to be my last response that takes a substantial amount of effort to compose. Arguing with /r/confidentlyincorrect-tier zero-effort Gish gallops* is not a good use of my time.
* especially from someone shameless about editing their posts after the fact to make them diverge substantially from what was originally written
> OpenAI keeps asking for copies, NYT keeps giving them out
The computer said ok isn’t enough or hacking would be legal.
> from an engineering standpoint it certainly sounds wrong
An engineering standpoint is completely irrelevant, this is a pure legal matter. The law is a strange place with it’s own rules you need to actually look into it not just make assumptions.
You seem confused. I didn't eat two tamales for lunch yesterday. I ate three. Don't make assumptions.
(Why do I need to post not just one but two exhortations to follow the basic rules of conversation? If you're going to say something, then say something true, but make sure it's also relevant.)
Hitting your partner is at the very least inconsiderate to them. Depending on your jurisdiction, it's probably also varying degrees of illegal. If you need some guidance on how to communicate without expressing it in the form of physical violence, I can try to help, but ultimately it's up to you.
The way that the law is treated by courts can be very unintuitive to programmers, who often seem to think that all it takes is to construct some clever bytecode sequence for CourtVM and then you win. It doesn't work like that. There may be someone who can explain in a way that you understand better, but the short version is that you are likely to be very disappointed by treating it like something that can be "hacked".
I was using existing precedent which doesn’t act like bite code but does weigh heavily, while you were referring to intuition. Such as my point that a computer responding with yes doesn’t override a prior statement that someone doesn’t have permission to use a system which literally has hundreds of examples and isn’t in doubt. You seemingly understand you lost, but keep wanting to respond.
Thus I accept your defeat but find your continued commentary silly.
It's almost as if posting condescending explanations that don't track the discussion while smugly framing it as if you're correcting some misapprehension of the person you're interacting with is something that can be obnoxious.
It's okay if you're still confused. The world is tricky. Not everyone is equipped to reason about everything they encounter, even they see other people are, and that's okay if you can't. But if you're going to offer your interpretation you need to actually look into what the other person is arguing—not just make assumptions.
A change in scale is a change in kind. There's a real qualitative difference between a party popper and a flashbang even though they're both "just" generators of light and sound. It's the same with regular ol' human learning---for which we have millenia of understanding and norms---and comparatively low-cost global-scale machine learning. The scale of the latter obliterates the structures we've built under the assumption of the former, so you can't just claim they're the same.
Our laws for carbon-based brains were designed pretty explicitly for carbon-based brains. Why assume these completely different things should be (legally) equivalent despite only superficial similarities?
I'd argue you're not looking very deep. On a physical level, chips and brains are entirely dissimilar. Some algorithms may share the same names, but that's about it. Brains aren't doing matrix multiplications all the time and chips don't use the Krebs cycle.
...Are you forgetting the fact that every clickwrapped application that doesn't have a network gated auth is just a binary patch away from being "legally impotent".
In point of fact, one of what I consider the "shadow innovations" of the technical era is that we even accepted the idea of clickwrapped contracting as a valid legal construct; especially given the fact a canonical requirement of a contract "a meeting of the minds" never substantially happens. (No redline accommodation, no valid contract imo).
Essentially, it may very well be the case that AI resynthesis is in a category so novel that copyright (which is itself a completely artificial construct composed wholly of law, with no grounding in natural rights whatsoever) has not predicted it and is utterly mute on what should be done with it one way or the other.
If that is the case, the time has come for novel law to address the issue. And that will be a hell of a political fight because it will basically boil down to who we feel should have more rights and resources in society.
YouTube is still based on copyright violation but there is no way to stop it so publishers play ball. There is no way music companies would allow their stuff on there for so cheap if they could stop it.
>> There is no way music companies would allow their stuff on there for so cheap if they could stop it.
Which is just another way of saying "Music Companies would stop technological progress itself if it affects their bottom line (and they can figure out a way to do it)."
they are now, yes. But that was not the case in the beginning, before Google bought it, people uploaded content completely without any regard to intellectual property and YouTube became the dominant streaming media platform and a couple of otherwise fairly skilled but hardly exceptional software engineers were made multi-millionaires) because it was the place to get free stuff that you'd otherwise have to pay for.
Yes and what we're talking about here is that the NYT is trying to get paid royalties as well, just like music & film publishers tried (successfully) to get Youtube to pay them. The long-term survival of ChatGPT may well depend on paying those royalties.
Did Microsoft use its vast source code repositories to rain its AI?
Did Google use its vast source code repositories to rain its AI ? They would also have vast s of user manuals etc. That are copyright? where these used too
Does anyone know ??
Copyright in its current form needs to die. It doesn't protect the artist, it protects their publishers and those who benefit from the sweat of their brow after the artist is dead. Protect works for 7 years renewable for 7 and no more.
7 years isn't realistic in the case of many authors and artists whose work becomes more valuable after time on the market, awards given as they mature etc.
Additionally if your concern is that artists and authors aren’t being paid, a 14 year copyright term is nothing. A corporation would just contract you to make a work, complete whatever minimum release is necessary to get copyright, then wait until it’s public domain to wide release a remastered version that they no longer have to pay royalties on since it is public domain.
Despite how many negatives the current copyright regime has, having to wait 70 years after the death of the author is far too long of a timeframe for that strategy to play out.
I'm not concerned about that. And if I was, why should artists be treated differently to anyone else who chooses to charge for the product of their efforts? Copyright didn't exist before 1710. Art sure did though. A protection of 7 years seems fine by that measure. Go make some more art after that expires. Or even before!
By secondary effects, do you mean artists would cease to create art? If so, then artists living in the many hundreds of years before copyright existed in law would disagree. Or do you mean there isn't enough music, or film or literature, etc. in the world? Spotify doesn't have an input problem as far as I know.
Please accept my apologies. I mean that the ostensible reason for copyright is to protect and thereby incentivise the production of artistic work. By extension, without copyright, artists would not create anything. But artistic work existed before copyright was created in law in 1710. Ergo, protection of 7 years would seem more than generous when, for example, builders are not paid every time somebody enters their buildings, or lawyers are not paid every time somebody reads the contracts they wrote.
If I install the plumbing in your house, you don't pay me every time you have a shower. I'm willing to concede an artist might be "better" than a plumber (commercially), but if copyright is intended to incentivise production of art, then 7 years seems perfectly reasonable time to let a work accrue some money before it goes into the public domain. 7+7 was in fact how copyright first existed.
Reading comments like yours convinces me that open sourcing my stuff wouldn't be a good idea. I very much like the idea that I own what I created. At least as long as owning stuff is how society works.
What is the value of all the software I've been able to use all my life for free? What is the value of the fact that all of that software behaved sanely and in my interests at all times? Both are basically incalculable. There is more than dollars to consider in the equation, and there are even many dollars too, merely indirectly.
> I think most people would agree that the physical trades and the arts are not well-suited for comparisons.
How is the writing of a book, the production of a movie or the recording of a song any different from the building of a bridge, the drafting of a contract, or the provision of management consultancy? How are they "fundamentally too different" if it's not the way we pay for them?
But more importantly, how is is that we see this difference as so fundamental that we afford protection to an artist (or more accurately, their publishers and the industry that surrounds them) via copyright for so long?
If the copyright system existed for shovels in that analogy, nobody else could make a shovel like yours during your lifetime, and for about 50 years after that, during which time your rights holders would get rich and shovel design would stand still.
Again - the copyright system as it currently exists needs to die. 7 years protection is enuogh.
The question was how is the value of the labor of an author different from the value of the labor of a plumber. There is a difference, and part of the difference is that if someone wants another sink installed tomorrow, they have to hire the plumber again to do another days work, while the authors work can be copied for free and, if allowed, a company will happily pay the author $50 one time, and then resell that same work 500 million times over for $50 each.
The point was not to defend the current copyright system and certainly not some of the more notable abusers. You asked a question, and that that is one part of the answer. Writing a book or doing any form of creative information-product work is different from doing physical-product or service work.
Thats not really relevant unless you think it will significantly reduce high quality output, which i dont think it will. Copyright is a perversion of the free market as a trade off to incent fixed cost investment (i.e. creating content) in an industry with zero marginal cost (i.e distribution is so close to free it might as well be thought of as free). The goal should be to pervert the free market as little as possible because raising the distribution price above zero very clearly causes deadweight loss to society and the longer the copyright term the more tenuous the claim of increased productiin and venefit to society is relative to the large deadweight loss it causes.
> Copyright in its current form needs to die. It doesn't protect the artist, it protects their publishers
Then it isn't the copyright at fault here but the agreement between the artist/author and publisher. Artists are famously forced to make bad deals when they don't have any leverage.
Agree but also disagree on the terms. Make the renewal fee exponponential (i.e first 10 year term is included second term to n terms renewal costs 100^n, which would make the renewals cost 10k,1mm,100mm,10b, etc.) That would put the vast majority of content into the public domain after the first term and let the amazing stuff pay the author longer but not much longer.
Perhaps it’s time to end the notion of copyright entirely. It’s been flawed since its inception, but more obviously so since the digital age made it apparent that information wants to be free, and will always find a way to be so.
Does modern copyright have any purpose other than maintaining the existing class order?
87 comments
[ 8.0 ms ] story [ 229 ms ] threadMy guess is there’s going to be an exception given, but that’s not yet decided.
Changing the words spoken to sound sophisticated does not change anything.
Obligatory IANAL.
https://mashable.com/article/chatgpt-revealed-personal-data-...
It may not be conclusive, but it's highly suggestive:
> Amount and substantiality of the portion used in relation to the copyrighted work as a whole: Under this factor, courts look at both the quantity and quality of the copyrighted material that was used. If the use includes a large portion of the copyrighted work, fair use is less likely to be found; if the use employs only a small amount of copyrighted material, fair use is more likely.
> https://www.copyright.gov/fair-use/
Copyright is insidious because people tried all sorts of ways to get around it in the past. You can’t for example replace all the words in an article with synonyms and get a completely new work. Thus, asking ChatGPT to summarize an article is problematic because the output is based on the original work irrespective of what’s happening internally.
Summaries are explicitly permitted, which is why CliffNotes, Wikipedia, etc can publish synopses of copyrighted works.
Note how much of the ‘summery’ section isn’t just a listing of what happens but an interpretation of what happens: https://www.cliffsnotes.com/literature/f/fahrenheit-451/summ...
“At the same time, she also gives the reader the opportunity to see that the government has dramatically changed what its citizens perceive as their history. For example, Montag never knew that firemen used to fight actual fires...”
You are confusing training a model vs using it. This way a web browser can also be said to use NYT's content.
The point was people seem to assume only direct copies are problematic but that’s not the case. Training/feature extraction is just as mechanical as making an MP3.
Because you're trying build a silicon brain, and a carbon-based brain is allowed to learn from books and articles without permission (e.g. reading them at the library or online), so why shouldn't a silicon brain also be allowed to learn that way?
People should obtain their learning material legally. True statement (at least as "true" as any opinionated value judgement can be, I guess). The question is how, though, is that statement relevant here?
Even by the NYT's own telling (the way that OpenAI obtained their articles was from the NYT's website) what OpenAI did was not illegal copyright infringement. The problem is that NYT is under the impression that allowing search engines to access their paywalled stuff and hoping they don't do anything else with it besides putting up conventional, Google-style SERPs makes for copyright infringement if what one of them actually wants it for involves other reasons. It doesn't. There's an adequate legal instrument available for NYT to use if they want to enforce conditions on use: a contract. Do they have a contract that somebody violated? If not, they have no cause to go after anyone. And if they do have a contract that was violated, then that's still not copyright infringement—it's a breach of their contract.
If I open a store you have implied permission to enter the premises, but not stay inside for 16 hours. Further the second I ask you to leave you no longer have permission to be on the premises. The store doesn’t need to add a lock or put up signs, you’re aware you don’t have permission and that’s it.
(Your analogy is bad and doesn't hold up. Copyright doesn't grant rightsholders control of the sort required here. It grants them the right to make and distribute copies. It doesn't grant the right to undistribute copies when it turns out they don't like what someone is doing with them.)
A core argument OpenAI is making is transitory copying is allowed as long training is fair use. But a permanent copy stored in a training database isn’t transitory and would itself be copyright infringement, so they don’t do that.
Thus training each version requires permission to download a new copy, which they now lack.
> ChatGPT keeps redownloading works to avoid the issue of keeping permeant copy’s of the training material
I don't know enough about how ChatGPT works to know whether or not that's true, but from an engineering standpoint it certainly sounds wrong because of how insane it would be if true, and I'm not at all convinced that you're right about this given how poorly you understand the other stuff you're trying to argue that I happen to know you're wrong about, so it doesn't seem unwise to conclude that the same is probably true about your claims here. But it doesn't matter, anyway:
> which would be obvious copyright infringement
Wrong. OpenAI keeps asking for copies, NYT keeps giving them out (whether/despite OpenAI having/not having "permission" or not). Not copyright infringement, let alone "obvious copyright infringement".
This is going to be my last response that takes a substantial amount of effort to compose. Arguing with /r/confidentlyincorrect-tier zero-effort Gish gallops* is not a good use of my time.
* especially from someone shameless about editing their posts after the fact to make them diverge substantially from what was originally written
The computer said ok isn’t enough or hacking would be legal.
> from an engineering standpoint it certainly sounds wrong
An engineering standpoint is completely irrelevant, this is a pure legal matter. The law is a strange place with it’s own rules you need to actually look into it not just make assumptions.
(Why do I need to post not just one but two exhortations to follow the basic rules of conversation? If you're going to say something, then say something true, but make sure it's also relevant.)
If you don’t understand I can clarify, but being obtuse just makes you seem childish.
PS: If analogy is annoy you I could stop, but you don’t understand what’s involved so I am trying to help.
Thus I accept your defeat but find your continued commentary silly.
It's okay if you're still confused. The world is tricky. Not everyone is equipped to reason about everything they encounter, even they see other people are, and that's okay if you can't. But if you're going to offer your interpretation you need to actually look into what the other person is arguing—not just make assumptions.
In point of fact, one of what I consider the "shadow innovations" of the technical era is that we even accepted the idea of clickwrapped contracting as a valid legal construct; especially given the fact a canonical requirement of a contract "a meeting of the minds" never substantially happens. (No redline accommodation, no valid contract imo).
If that is the case, the time has come for novel law to address the issue. And that will be a hell of a political fight because it will basically boil down to who we feel should have more rights and resources in society.
Is it actually the public, or is it three dinosaurs wearing a PR campaign?
Which is just another way of saying "Music Companies would stop technological progress itself if it affects their bottom line (and they can figure out a way to do it)."
Despite how many negatives the current copyright regime has, having to wait 70 years after the death of the author is far too long of a timeframe for that strategy to play out.
In that case, by definition, the artist was paid.
Being paid for a spec or a per diem while making the movie can be peanuts. Big name actors and writers make percentages on the backend.
Artists would still make art but would have to either have patronage or find an alternative income source to make a living.
This seems utterly arbitrary. I fail to understand your logic here.
There. Fix't.
If you want to make money, open-sourcing your work seems to be a very hard way of doing so.
The artist has no sort of guarantee of income from any particular piece of work or really even subsistence from their entire life's output of work.
I think most people would agree that the physical trades and the arts are not well-suited for comparisons. They are just fundamentally too different.
How is the writing of a book, the production of a movie or the recording of a song any different from the building of a bridge, the drafting of a contract, or the provision of management consultancy? How are they "fundamentally too different" if it's not the way we pay for them?
But more importantly, how is is that we see this difference as so fundamental that we afford protection to an artist (or more accurately, their publishers and the industry that surrounds them) via copyright for so long?
You do not get to clone the shovel by magic for about .1 cent and sell 50M copies of it for $20 each for the next forever.
Again - the copyright system as it currently exists needs to die. 7 years protection is enuogh.
The point was not to defend the current copyright system and certainly not some of the more notable abusers. You asked a question, and that that is one part of the answer. Writing a book or doing any form of creative information-product work is different from doing physical-product or service work.
Then it isn't the copyright at fault here but the agreement between the artist/author and publisher. Artists are famously forced to make bad deals when they don't have any leverage.
No more software licenses?
- There's no such thing as New York Times.
Does modern copyright have any purpose other than maintaining the existing class order?