Are neural nets (the weights file) copyrightable?

[gonzochess75]

BTW, I've seen a ton of actual proprietary copyright licenses. Company foo licenses such and such software to company bar. The terms and conditions can be incredibly onerous and unreasonable. The more valuable the property that foo has the more onerous. In some cases, they make demands on future products, on company alliances, on non disclosure agreements that limit future activity... on jurisdiction to hear the case... incredible demands. The GPL is quaint and meek in comparison. And if bar doesn't meet those demands, then they simply don't have any rights to foo's product that are protected by copyright. Period.

What do you guys think that Disney's license to a third party for limited use of let's say Star Wars looks like? I can tell you it'd be basically equivalent to agreeing to give your first born to Disney. And if you don't meet those terms or a judge says the terms are illegal, then the copyright contract is not valid and all rights go back to the copyright holder. Bar does not automatically get granted some rights. Bar has no copyrights other than those granted by the copyright holder. Don't believe me? Have a look at some cases.

Can you please calm down. Nobody here is attacking the GPL. We are just trying to understand what it says and what it doesn't. Or what it can and cannot impose.

Fair. The above wasn't intended as an attack on anyone.

[syzygy]

All of these hypotheticals are missing the forest for the trees. If you accept that the NN weights file is copyrightable, then the only relevant question is whether the SF + NN weights file is a "combined work" or if it is two separate works merely aggregated together.

In other words: the question is if the NN weights file is covered by the SF copyright, i.e. whether it is a derived work. Clearly it is not, as it contains no material from SF.

https://www.gnu.org/licenses/old-licens ... ggregation

The FSF does not decide what constitutes a derived work. Copyright law does.

6) The fact that he had to modify SF sources at all is evidence they are a combined work. If they were really two separate works not intended to be combined, then why did he have to modify SF sources to make them work well together?

Intended to be combined is irrelevant.

If I write a replacement chapter for an existing book, that replacement chapter is not protected by the copyright on the book just because I intended the chapter to be combined with the remaining chapters of the book. The relevant question is: does the replacement chapter include copyrighted material from the existing book.

However, I agree with you that there is no copyright on the NN weights.

[Michel]

All of these hypotheticals are missing the forest for the trees. If you accept that the NN weights file is copyrightable, then the only relevant question is whether the SF + NN weights file is a "combined work" or if it is two separate works merely aggregated together.

In other words: the question is if the NN weights file is covered by the SF copyright, i.e. whether it is a derived work. Clearly it is not, as it contains no material from SF.

https://www.gnu.org/licenses/old-licens ... ggregation

The FSF does not decide what constitutes a derived work. Copyright law does.

6) The fact that he had to modify SF sources at all is evidence they are a combined work. If they were really two separate works not intended to be combined, then why did he have to modify SF sources to make them work well together?

Intended to be combined is irrelevant.

If I write a replacement chapter for an existing book, that replacement chapter is not protected by the copyright on the book just because I intended the chapter to be combined with the remaining chapters of the book. The relevant question is: does the replacement chapter include copyrighted material from the existing book.

However, I agree with you that there is no copyright on the NN weights.

The GPL does not affect the copyright of software bundled with it. It simply prohibits the distribution of GPLed software as part of a larger software package which in total is not GPL compatible.

[syzygy]

This is false! A dynamically linking executable contains no material from the library, so cannot be touched by the copyright on that library when it is distributed.

As I understand it you cannot distribute a GPL'ed program unless you abide by its license (obviously). The GPL prohibits distributing a GPL'ed program together with a program which does have a GPL incompatible license if the two together constitute a modification of the original program (so that the precise aggregation exception does not apply).

So ChessBase can obviously distribute the NNUE net under a non-freely distributable license (assuming this is possible for a NN), but it cannot distribute it with SF as a unit, since this goes against the license of SF.

Yes, this seems to be the most that can be achieved with a copyright license. In principle, I could release a program under a license that forbids any person distributing it to have blue eyes or to use Microsoft Windows (i..e, if you have blue eyes or use Microsoft Windows, you don't get my permission for distributing the program).

Chessbase can distribute a modified SF and it can distribute the NNUE net.

A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an “aggregate” if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate.

The last sentence is effectively stating that 1+1=2. There is no way that a license on one work can effectively declare that it also applies to other works. If we delete the last sentence, just a definition remains. No legal consequence is stated.

So I don't quite see that GPLv3 forbids Chessbase to distribute SF together with the NNUE net, even if NNUE net is copyrighted. (Unless they are combined in a way that creates a single "work", but that is determined by copyright law, not by what the FSF wrote in the GPLv3 or in the FAQ.)

Or am I missing some other relevant part of the GPLv3? (The above quote is from secton 5 "Conveying Modified Source Versions".)

[Michel]

This is false! A dynamically linking executable contains no material from the library, so cannot be touched by the copyright on that library when it is distributed.

As I understand it you cannot distribute a GPL'ed program unless you abide by its license (obviously). The GPL prohibits distributing a GPL'ed program together with a program which does have a GPL incompatible license if the two together constitute a modification of the original program (so that the precise aggregation exception does not apply).

So ChessBase can obviously distribute the NNUE net under a non-freely distributable license (assuming this is possible for a NN), but it cannot distribute it with SF as a unit, since this goes against the license of SF.

Yes, this seems to be the most that can be achieved with a copyright license. In principle, I could release a program under a license that forbids any person distributing it to have blue eyes or to use Microsoft Windows (i..e, if you have blue eyes or use Microsoft Windows, you don't get my permission for distributing the program).

Chessbase can distribute a modified SF and it can distribute the NNUE net.

A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an “aggregate” if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate.

The last sentence is effectively stating that 1+1=2. There is no way that a license on one work can effectively declare that it also applies to other works. If we delete the last sentence, just a definition remains. No legal consequence is stated.

So I don't quite see that GPLv3 forbids Chessbase to distribute SF together with the NNUE net, even if NNUE net is copyrighted. (Unless they are combined in a way that creates a single "work", but that is determined by copyright law, not by what the FSF wrote in the GPLv3 or in the FAQ.)

Or am I missing some other relevant part of the GPLv3? (The above quote is from secton 5 "Conveying Modified Source Versions".)

I was confused by this as well but the way it was explained to me is that the GPL just states conditions under which GPLed software can be distributed. If none of these conditions are satisfied then you simply cannot distribute it. I assume that this is a standard feature of copyright law.

[syzygy]

Now, the explicit language of the GPL defines what it considers to be a mere aggregation vs a combined work. You may not like its definitions, but that would be like complaining about the definitions of what foouser really means when he says to, "go #(@&#*(@&$ yourself." You could of course argue in a court of law that the definition is unclear and you went and gave your self oral pleasure so that should count and the judge might agree with you. That'd be up to the judge.

But that definition doesn't hurt me or anyone else, and it is not used in the GPLv3 to forbid anything or to limit any rights to distribute the program. Perhaps a drafting error, but that's not my problem.

[syzygy]

I was confused by this as well but the way it was explained to me is that the GPL just states conditions under which GPLed software can be distributed. If none of these conditions are satisfied then you simply cannot distribute it. I assume that this is a standard feature of copyright law.

By reading more of your posts I realised we agree very much on the "aggregate" section.

But I don't see where the GPL, which gives me the right to distribute SF, states that I cannot do so as part of a "non-aggregate" compilation. (It is different if SF is combined with the other parts to become one "work", but I don't see that happening if SF and the NNUE net are combined in one zip file, for example. If they would form one "work", then that work is protected by SF's copyright (and by the copyright on the NNUE net, if any).)

[gonzochess75]

Now, the explicit language of the GPL defines what it considers to be a mere aggregation vs a combined work. You may not like its definitions, but that would be like complaining about the definitions of what foouser really means when he says to, "go #(@&#*(@&$ yourself." You could of course argue in a court of law that the definition is unclear and you went and gave your self oral pleasure so that should count and the judge might agree with you. That'd be up to the judge.

But that definition doesn't hurt me or anyone else, and it is not used in the GPLv3 to forbid anything or to limit any rights to distribute the program. Perhaps a drafting error, but that's not my problem.

You have it right when you say this, 'In principle, I could release a program under a license that forbids any person distributing it to have blue eyes or to use Microsoft Windows (i..e, if you have blue eyes or use Microsoft Windows, you don't get my permission for distributing the program).'

That's essentially what the GPL is doing. It has a definition of aggregate vs combined work. If you don't meet its definition of aggregate, then you don't get to distribute the work covered by the GPL unless you adhere to its terms. You get to choose whether you want to meet those terms or not. However, if you don't then you have no rights to the work in question.

[gonzochess75]

I was confused by this as well but the way it was explained to me is that the GPL just states conditions under which GPLed software can be distributed. If none of these conditions are satisfied then you simply cannot distribute it. I assume that this is a standard feature of copyright law.

This is exactly right as far as I understand.

[syzygy]

That's essentially what the GPL is doing. It has a definition of aggregate vs combined work. If you don't meet its definition of aggregate, then you don't get to distribute the work covered by the GPL unless you adhere to its terms. You get to choose whether you want to meet those terms or not. However, if you don't then you have no rights to the work in question.

Where does it say that if you don't meet its definition of aggregate, then you don't get to distribute the GPL-protected work? I don't see it. Am I overlooking it?