Are neural nets (the weights file) copyrightable?

[syzygy]

But a reviewer such as Lawrence Rosen seems to agree with me that it is fine to package independent copyrightable works together even if they combine to make something more sophisticated.

It does seem Lawrence Rosen agrees with you, but his comments are in the context of the GPLv2 and not the GPLv3 right?

No, this is about GPLv3, draft 4, which I don't think differs on this point from GPLv3, final. (I included a link to draft 4.)

[Michel]

GPLv3 is indeed worded differently, but this shows that whatever is in the FSF's GPL FAQ has to be taken with a big grain of salt.

I assume that because of the ambiguities in the GPL 2 they decided to get serious legal council for the GPL 3. By using their own definition of "covered work" they circumvented having to rely on copyright law to define a derived work.

A covered work is unmodified SF or "a work based on SF", which is the modified FF2 SF.

This not how I see it. For me FF2 (SF+net) is the “work based on SF”. This the common sense interpretation of what the license says. Anyway this can only be decided in court I guess.

[towforce]

I want to make it clear that I do not know if a weights file is copyrightable or not.
I am not at all sure either way on that particular thing.
But I am sure that data is not part of the source code.

It is possible that I misunderstood your motivation. But I find it very, very strange that someone who knows some computer science pretends that source code means "source code + data" unless they are trying to twist something to their own ends. And without some sort of motivation I do not know why someone would do that.

So my question is, "Why are you doing that?"

I profoundly disagree with the statement that "data is not part of the source code" in relation to an NN weights file. IMO, it is source code: it tells the program how to do its calculations.

A database can easily be part of an application, and if the database and data are bundled in with a published GPL application, then everyone would be entitled to see that data.

Of course, a software author would not be allowed to publish personal data (e.g. the UK's Data Protection Act).

[syzygy]

GPLv3 is indeed worded differently, but this shows that whatever is in the FSF's GPL FAQ has to be taken with a big grain of salt.

I assume that because of the ambiguities in the GPL 2 they decided to get serious legal council for the GPL 3. By using their own definition of "covered work" they circumvented having to rely on copyright law to define a derived work.

A covered work is unmodified SF or "a work based on SF", which is the modified FF2 SF.

This not how I see it. For me FF2 (SF+net) is the “work based on SF”. This the common sense interpretation of what the license says. Anyway this can only be decided in court I guess.

But I want to distribute the FF2 SF executable, and this is certainly an "entire" covered work. (Note that an "entire" work can even be completely disfunctional, which happens all the time on github.)

The zip file bundles two separate works, I don't need special copyright permisson for distributing the zip file as a whole. It was me who bundled them, so any copyright on the creating the "collection" is mine (but obviously there is no creativity there).

I don't care what Stallman and Moglen want the GPLv3 to achieve. If it were up to them, a program dynamically linking to a GPL'd library is protected by the copyright on the library. They are free to dream...

[Michel]

We can argue forever. From my POV by bundling SF with a net, CB has created a work based on SF (that's what they are admitting themselves). So in GPL 3 terminology it is a "covered work". Somewhat confusingly this does not mean that the bundle automatically falls under the GPL 3 since the aggregation exception might apply. Assuming it doesn't (I am relying on the GPL FAQ for this) then the bundle falls under the GPL 3 and CB is indeed committing a GPL violation by distributing it (assuming the net is under a non free license).

So it all depends on the definition of "a work based on SF". I am just following the common sense interpretation of what these words mean. Ultimately a license is written in natural language and in the absence of explicit redefinitions one has to use the common meaning of the words.

[syzygy]

We can argue forever. From my POV by bundling SF with a net, CB has created a work based on SF (that's what they are admitting themselves). So in GPL 3 terminology it is a "covered work". Somewhat confusingly this does not mean that the bundle automatically falls under the GPL 3 since the aggregation exception might apply. Assuming it doesn't (I am relying on the GPL FAQ for this) then the bundle falls under the GPL 3 and CB is indeed committing a GPL violation by distributing it (assuming the net is under a non free license).

So it all depends on the definition of "a work based on SF". I am just following the common sense interpretation of what these words mean. Ultimately a license is written in natural language and in the absence of explicit redefinitions one has to use the common meaning of the words.

I'm asking myself which permissions copyright law requires me to obtain to distribute the bundle of FF2 SF and FF2 NNUE (if I were Chessbase and FF2 NNUE was copyrighted). The "collective" work is mine, as I created the bundle (see here for US copyright law, it is not different in the EU). FF2 NNUE is mine. I only need permission to distribute the work FF2 SF. I therefore look at the GPLv3 to see whether I have permission to distribute the work FF2 SF (as part of the bundle). So the "covered work" is FF2 SF. I agree that, in principle, a license could specify that a covered work may not be distributed by a person with blue eyes or in a ZIP file with anything not licensed under XYZ, but the GPLv3, on close inspection, does not seem to specify such a thing.

[Michel]

We can argue forever. From my POV by bundling SF with a net, CB has created a work based on SF (that's what they are admitting themselves). So in GPL 3 terminology it is a "covered work". Somewhat confusingly this does not mean that the bundle automatically falls under the GPL 3 since the aggregation exception might apply. Assuming it doesn't (I am relying on the GPL FAQ for this) then the bundle falls under the GPL 3 and CB is indeed committing a GPL violation by distributing it (assuming the net is under a non free license).

So it all depends on the definition of "a work based on SF". I am just following the common sense interpretation of what these words mean. Ultimately a license is written in natural language and in the absence of explicit redefinitions one has to use the common meaning of the words.

I'm asking myself which permissions copyright law requires me to obtain to distribute the bundle of FF2 SF and FF2 NNUE (if I were Chessbase and FF2 NNUE was copyrighted). The "collective" work is mine, as I created the bundle (see here for US copyright law, it is not different in the EU). FF2 NNUE is mine. I only need permission to distribute the work FF2 SF. I therefore look at the GPLv3 to see whether I have permission to distribute the work FF2 SF (as part of the bundle). So the "covered work" is FF2 SF. I agree that, in principle, a license could specify that a covered work may not be distributed by a person with blue eyes or in a ZIP file with anything not licensed under XYZ, but the GPLv3, on close inspection, does not seem to specify such a thing.

On what grounds do you not consider FF2 (=SF+net) to be a work based on SF?

This has nothing to do with copyright law. It is just common English. In the absence of redefinitions words should be assumed to have their common meaning.

EDIT: Fixed double negation issue.

EDIT2: Note that to define a "modification" the GPL 3 invokes copyright law. But for the definition of a "covered work" it does not.

[chrisw]

I want to make it clear that I do not know if a weights file is copyrightable or not.
I am not at all sure either way on that particular thing.
But I am sure that data is not part of the source code.

It is possible that I misunderstood your motivation. But I find it very, very strange that someone who knows some computer science pretends that source code means "source code + data" unless they are trying to twist something to their own ends. And without some sort of motivation I do not know why someone would do that.

So my question is, "Why are you doing that?"

I profoundly disagree with the statement that "data is not part of the source code" in relation to an NN weights file. IMO, it is source code: it tells the program how to do its calculations.

Nonsense.

A database can easily be part of an application, and if the database and data are bundled in with a published GPL application, then everyone would be entitled to see that data.

Of course, a software author would not be allowed to publish personal data (e.g. the UK's Data Protection Act).

[syzygy]

On what grounds do you not consider FF2 (=SF+net) to be a work based on SF?

This has nothing to do with copyright law. It is just common English. In the absence of redefinitions words should be assumed to have their common meaning.

Everything starts with the copyright law question which permissions I need. The permission I need is the permission to distribute FF2 SF. So I invoke the GPLv3 to obtain permission for distributing the entire covered work "FF2 SF" (as part of a zip file), not for distributing the zip file as a whole (of which only FF2 SF is protected by the SF copyright).

The wording of the GPLv3 may allow one to consider the zip file to be a "work based on SF", but I have already covered my copyright needs by invoking the GPLv3 with FF2 SF as the "work based on SF".

And certainly it makes more sense to invoke the GPLv3 on FF2 SF, as this is the derived work that undisputably is protected by the SF copyright. The zip file itself is not a derived work, it merely contains one. (And there are commentators suggesting that "work based on SF" is nothing else than "derived work".)

[Michel]

On what grounds do you not consider FF2 (=SF+net) to be a work based on SF?

This has nothing to do with copyright law. It is just common English. In the absence of redefinitions words should be assumed to have their common meaning.

Everything starts with the copyright law question which permissions I need. The permission I need is the permission to distribute FF2 SF. So I invoke the GPLv3 to obtain permission for distributing the entire covered work "FF2 SF" (as part of a zip file), not for distributing the zip file as a whole (of which only FF2 SF is protected by the SF copyright).

The wording of the GPLv3 may allow one to consider the zip file to be a "work based on SF", but I have already covered my copyright needs by invoking the GPLv3 with FF2 SF as the "work based on SF".

And certainly it makes more sense to invoke the GPLv3 on FF2 SF, as this is the derived work that undisputably is protected by the SF copyright. The zip file itself is not a derived work, it merely contains one. (And there are commentators suggesting that "work based on SF" is nothing else than "derived work".)

SF is certainly a work based on SF. But so is FF2. The GPL 3 wants to put restrictions on both of them. With SF there is tautologically no problem but with the work FF2 there may be.