Fat Fritz 2

[gaard]

In fact, on further consideration, this argument doesn't make any sense (I'll admit there could be something I'm not understanding).

Here's the situation as I understand it:

ChessBase is conveying a binary under the GPLv3 license. Therefore, as a recipient of the binary, I have the freedom to do whatever I want with it, provided I both provide a source and attach a GPLv3 license to all those I convey it to in turn.

Your argument, in my understanding, is that some third party might have licensed the parameters included in the binary ChessBase is distributing under a different, GPLv3 incompatible, license. Therefore, ChessBase is illegally distributing the binary. Furthermore, I, as the recipient of the GPLv3 licensed binary from ChessBase or those having indirectly received the binary from ChessBase, would be at fault, legally, were I to extract the parameters from the binary.

This doesn't make any sense for two reasons:

1.
In the above description of the situation, I would not be at fault at ChessBase would be responsible for all damages incurred on said third party.

2.
It is not possible to license numbers.

Let me guess, you’re not a lawyer?

The poster already said he was not a legal expert. Surely you know of another way to deflect and distract than attacking the messenger?

[dkappe]

1.
In the above description of the situation, I would not be at fault at ChessBase would be responsible for all damages incurred on said third party.

2.
It is not possible to license numbers.

Correct Neither are you, though as you are speaking from a position of such authority on the subject, I would appreciate if you could explain how the above is incorrect.

1. You’re asking whether I would be solely liable for damages you committed under the false belief that Oracle’s software was covered by the GPLv3? I don’t know. My suspicion is that Oracle would sue the daylights out of both of us. I do think this is well beyond the scope of the GPL, however.

2. Maybe a better question is, can pretrained deep learning models be copyrighted or licensed? The answer is, maybe. If the answer is yes, then both the stockfish developers and whoever owns FF2 are ok. If not, then neither is OK. For example both GPLv3 and CC0 would be meaningless as far as the stockfish nets are concerned. Most likely the litigant with the deepest pockets wins.

[connor_mcmonigle]

1.
In the above description of the situation, I would not be at fault at ChessBase would be responsible for all damages incurred on said third party.

2.
It is not possible to license numbers.

Correct Neither are you, though as you are speaking from a position of such authority on the subject, I would appreciate if you could explain how the above is incorrect.

1. You’re asking whether I would be solely liable for damages you committed under the false belief that Oracle’s software was covered by the GPLv3? I don’t know. My suspicion is that Oracle would sue the daylights out of both of us. I do think this is well beyond the scope of the GPL, however.

2. Maybe a better question is, can pretrained deep learning models be copyrighted or licensed? The answer is, maybe. If the answer is yes, then both the stockfish developers and whoever owns FF2 are ok. If not, then neither is OK. For example both GPLv3 and CC0 would be meaningless as far as the stockfish nets are concerned. Most likely the litigant with the deepest pockets wins.

Thanks for providing an explanation of the thinking here.

1. I guess you're right that the protections of the GPL wouldn't necessarily prevent the recipient of the conveyed software from being liable for the damages. However, you do have to admit that it's a rather comical situation you're describing: ChessBase's distributed GPLv3 binary is (maybe) protected from free redistribution/modification as their distribution of that binary is illegal (as it maybe contains components with incompatible licenses). It's only this uncertainty that makes selling this product profitable. If the distribution is determined to be illegal, then the product could no longer be sold by ChessBase. If the distribution is determined to be legal, then it can be freely redistributed and no one would purchase the product from ChessBase.

2. The situation is much murkier than I initially understood it to be as there doesn't appear to be any real precedent. Long term, I'd imagine network parameters are established as protected by a form of intellectual property, though there doesn't appear to be any clarity here.

[Modern Times]

Well, Chessbase would have not been in any doubt how controversial this would be and how some people might react to it. So they would surely have got legal opinions on it, either from in-house staff if they have any, or externally. So they must think they are on solid ground here. If they didn't check the legal side, well then that is a bit stupid.

[AndrewGrant]

A question back to Dkappe -- Is it a violation of (whomever; Albert?) trained the Network to reverse engineer the binary and extract the weights into a typical Stockfish form? To distribute that?

I would suggest anyone extracting or downloading then change some weights by a single RL run and then release it for free, . call it "Slim Shady Fritz", and then let see CB lawyers try to "prove" these new weights net infringe their "copyright" -IPs of the net...

Similar thinking here as the excellent work with the Rocket (houndini) engine.. "Slim Fritoz 2"

Well, inciting others to break the law doesn’t strike me as very clever.

I don’t know that FF2 behaves the same way, but it’s very easy to detect that Night Nurse is trained from mcts data rather than ab data. Also, RL training a mcts trained net will make it weaker in a hurry.

So good luck with that...from prison.

Aha, so you've taken a stance on something.

[dkappe]

Aha, so you've taken a stance on something.

If so I’ve made a terrible mistake.

[AndrewGrant]

Aha, so you've taken a stance on something.

If so I’ve made a terrible mistake.

Your comment implies that you view extracting the weights from Fat Fritz and providing them online for free it a violation of the "authors" of Fat Fritz. I believe this is the only time you have ever claimed something to be a violation. Otherwise you just jerk around saying everything under the sun is legal. Big mistake buddy, we've got you now.

[intrepid]

In the new article, "Fat Fritz 2: The Best of Both Worlds," Chessbase makes the following acknowledgement: "Also, warmest thanks to my friend Dietrich Kappe who is an endless source of useful scripts and ideas, and who helped take the project to the next level. His friendship and generosity cannot be overstated." Chessbase clearly failed to acknowledged dkappe's heroic "disabusing" of people on talkchess.

Other than that, Chessbase has reached another low with this FF2 release. They never report on computer chess, except to push their product, regardless of its inferiority or shady past (like Houdini). They fail to report on revolutionary advances like NNUE, instead making the big reveal after copy-pasting and rebranding Stockfish. They then compare this Stockfish clone to an old version of Stockfish, and top it all off by preventing people from making comments on their "articles" (promotional propaganda pieces) about FF2. Complete disgrace.

[dkappe]

Aha, so you've taken a stance on something.

If so I’ve made a terrible mistake.

Your comment implies that you view extracting the weights from Fat Fritz and providing them online for free it a violation of the "authors" of Fat Fritz. I believe this is the only time you have ever claimed something to be a violation. Otherwise you just jerk around saying everything under the sun is legal. Big mistake buddy, we've got you now.

A violation of the authors? Sounds unpleasant. You mean a violation of their rights under copyright law? Possibly. If neural network models can be copyrighted.

[dkappe]

In the new article, "Fat Fritz 2: The Best of Both Worlds," Chessbase makes the following acknowledgement: "Also, warmest thanks to my friend Dietrich Kappe who is an endless source of useful scripts and ideas, and who helped take the project to the next level. His friendship and generosity cannot be overstated." Chessbase clearly failed to acknowledged dkappe's heroic "disabusing" of people on talkchess.

Other than that, Chessbase has reached another low with this FF2 release. They never report on computer chess, except to push their product, regardless of its inferiority or shady past (like Houdini). They fail to report on revolutionary advances like NNUE, instead making the big reveal after copy-pasting and rebranding Stockfish. They then compare this Stockfish clone to an old version of Stockfish, and top it all off by preventing people from making comments on their "articles" (promotional propaganda pieces) about FF2. Complete disgrace.

Two areas of ignorance infuriate me: the GPL and the scientific method. Not sure that my methods make me a great PR asset.