Are neural nets (the weights file) copyrightable?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
gonzochess75
Posts: 190
Joined: Mon Dec 10, 2018 2:29 pm
Full name: Adam Treat

Re: Are neural nets (the weights file) copyrightable?

Post by gonzochess75 » Mon Feb 22, 2021 8:26 pm

dkappe wrote:
Mon Feb 22, 2021 8:15 pm
cpeters wrote:
Mon Feb 22, 2021 8:02 pm
dkappe/hello Dietrich,


is there a reason why you do you refrain from attaching any license to your own nets (like from https://github.com/dkappe/leela-chess-w ... d-Networks)?

As githubs terms and services/faq gives this:

https://docs.github.com/en/github/creat ... repository

am I not wrong to think it's more flexible for you (decide on a case-by-case basis) or would it be simply not worth the time/effort pondering the appropriate one, no?

Sometimes a missing license can be just an oversight*:
https://github.com/dkappe/TogaIII (as this should be GPL surely).

*edit: one fork (just this) wahtever one wants

greetings
Up until all this nonsense started, I wasn’t really focused on it.

The toga nets are intended to be embedded in the Toga III engine and are thus distributed under the GPL compatible CC0 license (I’ll make that explicit with the next release). Going forward, I’m releasing my other networks under the CC BY ND 4.0 license. https://creativecommons.org/licenses/by-nd/4.0/
AFAIK, that by-nd license is not GPL compatible as it provides additional restrictions above and beyond GPL. https://www.gnu.org/licenses/license-li ... onLicenses

cpeters
Posts: 15
Joined: Wed Feb 17, 2021 6:44 pm
Full name: Christian Petersen

Re: Are neural nets (the weights file) copyrightable?

Post by cpeters » Mon Feb 22, 2021 8:31 pm

I too think this Toga is GPL too, really. But I didn't look closer..

thanks anyway

kerfuffle!

dkappe
Posts: 855
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Are neural nets (the weights file) copyrightable?

Post by dkappe » Mon Feb 22, 2021 8:35 pm

gonzochess75 wrote:
Mon Feb 22, 2021 8:26 pm
AFAIK, that by-nd license is not GPL compatible as it provides additional restrictions above and beyond GPL. https://www.gnu.org/licenses/license-li ... onLicenses
Yep. If you wanted to embed it in a GPL licensed engine, you could not distribute it, but you could distribute it alongside with its own LICENSE.
Last edited by dkappe on Mon Feb 22, 2021 8:51 pm, edited 2 times in total.

cpeters
Posts: 15
Joined: Wed Feb 17, 2021 6:44 pm
Full name: Christian Petersen

Re: Are neural nets (the weights file) copyrightable?

Post by cpeters » Mon Feb 22, 2021 8:42 pm

"AFAIK, that by-nd license is not GPL compatible as it provides additional restrictions above and beyond GPL. "


True that; but does it matter in this case (if they aren't distributed in one binary)?

edit:
"Yep. If you wanted to embed it in a GPL licensed engine, you could distribute it."


I'm out/lost oversight.

Ckappe
Posts: 64
Joined: Sun Feb 14, 2021 10:50 am
Full name: Rütger Andersen

Re: Are neural nets (the weights file) copyrightable?

Post by Ckappe » Mon Feb 22, 2021 8:58 pm

dkappe wrote:
Mon Feb 22, 2021 7:50 pm
“Ckappe?” Are we related, or is that a passive aggressive handle on your account? :lol:
I don't think we are related, and I don't think neither your handle nor mine could be considered aggressive.
Last edited by Ckappe on Mon Feb 22, 2021 9:05 pm, edited 1 time in total.

dkappe
Posts: 855
Joined: Tue Aug 21, 2018 5:52 pm
Full name: Dietrich Kappe

Re: Are neural nets (the weights file) copyrightable?

Post by dkappe » Mon Feb 22, 2021 9:00 pm

Ckappe wrote:
Mon Feb 22, 2021 8:58 pm
dkappe wrote:
Mon Feb 22, 2021 7:50 pm
“Ckappe?” Are we related, or is that a passive aggressive handle on your account? :lol:
I don't think we are related.
Thanks for the confirmation, Mr. Rytger Andersen. I assume “Ckappe” is just some sort of nickname. :lol:

Ckappe
Posts: 64
Joined: Sun Feb 14, 2021 10:50 am
Full name: Rütger Andersen

Re: Are neural nets (the weights file) copyrightable?

Post by Ckappe » Mon Feb 22, 2021 9:07 pm

dkappe wrote:
Mon Feb 22, 2021 9:00 pm
Thanks for the confirmation, Mr. Rytger Andersen. I assume “Ckappe” is just some sort of nickname. :lol:
I guess we are straying off topic here...

Dann Corbit
Posts: 12012
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Are neural nets (the weights file) copyrightable?

Post by Dann Corbit » Tue Feb 23, 2021 7:55 am

As to the original question, "Are neural nets (the weights file) copyrightable?" that will depend on originality.
If, for instance, you simply launch a bevy of engines who labor for a few days and write the net, there is no originality in that.
If, on the other hand, you do lots of manual things (perhaps blending games from correspondence championships with games from TCEC and adding in specific test problems and endgame knowledge manually) then it might be copyrightable.

But I could be wrong about that. I think copyright issues are best resolved by people who know what they are doing, which means lawyers who have a degree in intellectual property laws

I guess that we have none of those in this forum/
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

AndrewGrant
Posts: 1030
Joined: Tue Apr 19, 2016 4:08 am
Location: U.S.A
Full name: Andrew Grant
Contact:

Re: Are neural nets (the weights file) copyrightable?

Post by AndrewGrant » Tue Feb 23, 2021 8:00 am

Dann Corbit wrote:
Tue Feb 23, 2021 7:55 am
As to the original question, "Are neural nets (the weights file) copyrightable?" that will depend on originality.
If, for instance, you simply launch a bevy of engines who labor for a few days and write the net, there is no originality in that.
If, on the other hand, you do lots of manual things (perhaps blending games from correspondence championships with games from TCEC and adding in specific test problems and endgame knowledge manually) then it might be copyrightable.

But I could be wrong about that. I think copyright issues are best resolved by people who know what they are doing, which means lawyers who have a degree in intellectual property laws

I guess that we have none of those in this forum/
I disagree with the premise in your first part wholly. Who is to say that hand selecting games is a more copyrightable process of building a network than the selection of the paramaters used to launch the fleet of engines?

This sort of vague idea that "Oh, my nets special because I want and hand picked out games from U13 mongolio", is the thing that Alberto Plata did at the start with DeusX; tricking the admins into thinking he had really gone above and beyond to build something special.

User avatar
hgm
Posts: 25862
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Are neural nets (the weights file) copyrightable?

Post by hgm » Tue Feb 23, 2021 8:30 am

I don't think it is contested in any way that individual Chess games are not copyrightable (in Europe), but that when a collection of chess games selected is published as a book, that book will be protected by copyright. Purely mechanical translation of the collection into another format will not destroy existing copyrights; it will just fail to add any new ('translation') copyrights, but inherit the copyrights from the collection.

So I think that Dann is right, and whether a neural net is copyrightable in principle depends on whether the training material offered to it in the learning process was copyrightable. This depends on the human effort spent in this process. Downloading a large database of games, and feeding it to existing NN training software obviously would not qualify.

I would imagine that the burdon of proof here is on the party that is challenging the copyrights. Because the mere fact that he seeks to copy the thing is strong evidence that it actually is valuable, and not just the result of some insignificant effort that he could easily duplicate himself without generating copyrights on the result. (So that it could never be a copyrights infringement.)

BTW, I don't understand the hassle. It seems that the FF2 net is not stronger than existing free NNUE nest at all, and that the claim of superiority of FF2 purely depended on using Stockfish-dev as a 'network player', and comparing it to Stockfish 12. So who cares whether the FF2 net is actually protected by copyrights or not? No one in his right mind should want to copy it. If ChessBase wants to dupe customers by selling them inferior stuff to what is obtainable for free, that is their business. (Hasn't that always been their core business?) The best the computer chess community could do to protect the potential victims of such a scam is to make sure the inferiority of FF2 is widely known, by prominently featuring it on the rating lists...

Post Reply