Stockfish: Our lawsuit against ChessBase

[chessica]

On the positive side, there's always the practical issue: if I asked an LLM to write a story, it's possible that nobody would know that I'd done that (although there are detection programs - just as there are programs that can differentiate human games from computer games (but not yet with 100% accuracy)). If nobody knew, then I could get away with claiming the copyright.

Acting as a criminal does not justify criminality if you get away with it.

Breach of copyright is a civil offence. There are circumstances under which it becomes a criminal offence - but they are very rare.

Under this scenario:

1. I create a copyrighted work

2. Some of that work was actually written by a chatbot

3. Another person copies some of that work

4. I sue that person for breach of copyright

If the other person could prove that some of my work was written by a chatbot, then they would win because the copyright would not then apply.

However, there would be no question of criminal law being used in this case - only civil law.

In which country, please? In Germany, the rule of law has been undermined. Only those who can pay the legal costs get justice. Just like in the USA.

[chesskobra]

Since this thread has now turned into copyright, AI etc. let me ask a question that I have been meaning to ask. What do you all think about putting your work on GitHub? In the past I uploaded a couple of repositories that have content that is more than just implementations of standard algorithms. But some time ago I decided not to upload any new work to GitHub, especially if it involves implementing research ideas and algorithms. I just do not trust that company. What about email? If you discuss research or any other sensitive content over email with your collaborators, can that content be used for training AI? FB is now planning to train AI on users' messages, photos, etc. Of course, you should not discuss major trade secrets over hotmail, but many of us are not discussing serious trade secrets but just our day to day research. How paranoid should we be?

I look forward to FSF or some other entity creating a very restrictive license under which your code cannot be fed into any AI training system. Has anybody attempted to create such a license?

Knuth has said that his work computer is never connected to the internet. I have not reached his level (neither in terms of the importance of my work nor in terms of protecting it). But I mostly use USB to to copy files from one computer to another. I have a Samsung tablet. I have not read in detail the terms and conditions, but it seems that they can do anything they want with the files on my device, even if I do not use their cloud service. Is that true? Also, are our devices only rented to us?

[towforce]

On the positive side, there's always the practical issue: if I asked an LLM to write a story, it's possible that nobody would know that I'd done that (although there are detection programs - just as there are programs that can differentiate human games from computer games (but not yet with 100% accuracy)). If nobody knew, then I could get away with claiming the copyright.

Acting as a criminal does not justify criminality if you get away with it.

Breach of copyright is a civil offence. There are circumstances under which it becomes a criminal offence - but they are very rare.

Under this scenario:

1. I create a copyrighted work

2. Some of that work was actually written by a chatbot

3. Another person copies some of that work

4. I sue that person for breach of copyright

If the other person could prove that some of my work was written by a chatbot, then they would win because the copyright would not then apply.

However, there would be no question of criminal law being used in this case - only civil law.

In which country, please? In Germany, the rule of law has been undermined. Only those who can pay the legal costs get justice. Just like in the USA.

This is both off-topic and a stretch of my expertise, but in Britain you can sue somebody under the "Small Claims Rules": if you do that, the most you can claim is £10,000.

Maybe we could use AI to take this further and introduce "Micro Justice".

[towforce]

Since this thread has now turned into copyright, AI etc. let me ask a question that I have been meaning to ask. What do you all think about putting your work on GitHub? In the past I uploaded a couple of repositories that have content that is more than just implementations of standard algorithms. But some time ago I decided not to upload any new work to GitHub, especially if it involves implementing research ideas and algorithms. I just do not trust that company. What about email? If you discuss research or any other sensitive content over email with your collaborators, can that content be used for training AI? FB is now planning to train AI on users' messages, photos, etc. Of course, you should not discuss major trade secrets over hotmail, but many of us are not discussing serious trade secrets but just our day to day research. How paranoid should we be?

I look forward to FSF or some other entity creating a very restrictive license under which your code cannot be fed into any AI training system. Has anybody attempted to create such a license?

Knuth has said that his work computer is never connected to the internet. I have not reached his level (neither in terms of the importance of my work nor in terms of protecting it). But I mostly use USB to to copy files from one computer to another. I have a Samsung tablet. I have not read in detail the terms and conditions, but it seems that they can do anything they want with the files on my device, even if I do not use their cloud service. Is that true? Also, are our devices only rented to us?

Hmmmm... it's not clear to me why people are worried about their code being used to train a chatbot: it seems to me that the probability of it making their lives significantly worse is vanishingly small.

If, say, you've written an algorithm that cracks cryptocurrency encryption, the description clearly states what it does, and you don't want other people being able to do it, then maybe don't put it on a source-control host.

[chesskobra]

Hmmmm... it's not clear to me why people are worried about their code being used to train a chatbot: it seems to me that the probability of it making their lives significantly worse is vanishingly small.

You could say the same thing about GPL violation in many cases. If Stockfish license is violated, whose life is significantly worsened? The point is why should a commercial entity use someone's work without credit? I am happy to share my code in most cases, just not with a commercial entity. "Your private repositories will not be looked at by a human" - there you go. Now on, if I were learning a new language and implemented tic-tac-toe, the code would go to GitHub for the benefit of society.

[smatovic]

...this thread derailed meanwhile into AI and copyright++ law in general, so:

Nvidia Denies Pirate e-Book Sites Are 'Shadow Libraries' To Shut Down Lawsuit
https://yro.slashdot.org/story/24/05/28 ... wn-lawsuit
[...]

Hmm, books are fair use, and now web pages are freeware? Content industry vs. AI industry.

Microsoft's AI CEO: Web Content (Without a Robots.txt File) is 'Freeware' for AI Training
https://slashdot.org/story/24/07/07/005 ... i-training

[...]
"Yeah, I think — look, it's a very fair argument." SULEYMAN: "I think that with respect to content that is already on the open web, the social contract of that content since the 90s has been that it is fair use. Anyone can copy it, recreate with it, reproduce with it. That has been freeware, if you like. That's been the understanding. "There's a separate category where a website or a publisher or a news organization had explicitly said, 'Do not scrape or crawl me for any other reason than indexing me so that other people can find that content.' That's a gray area and I think that's going to work its way through the courts."
[...]

As others mentioned, we might miss an open source code and data license to explicitly drop out of AI training stuff resp. for open source neural networks only. Or the courts decide how to interpret open source code licenses for training commercial neural networks.

--
Srdja

[syzygy]

I look forward to FSF or some other entity creating a very restrictive license under which your code cannot be fed into any AI training system. Has anybody attempted to create such a license?

Based on what? Copyright cannot be used to prevent others from training their own brain or training their AI system on your code.

[Ras]

Copyright cannot be used to prevent others from training their own brain or training their AI system on your code.

Of course it can - as derivative work. That's why AI companies are striking deals with news corporations. Code is no different from other text when it comes to copyright. Whether or not that training actually qualifies as derivative work, or whether the point of the deals is that they are cheaper than lawsuits for answering that question, is still an open point. But it's not like there's no legal basis at all.

[syzygy]

Copyright cannot be used to prevent others from training their own brain or training their AI system on your code.

Of course it can - as derivative work. That's why AI companies are striking deals with news corporations. Code is no different from other text when it comes to copyright. Whether or not that training actually qualifies as derivative work, or whether the point of the deals is that they are cheaper than lawsuits for answering that question, is still an open point. But it's not like there's no legal basis at all.

A derivative work under copyright law needs to contain creative expression from the original work. This is not the case when the original work is only used for adjusting weights. The legal situation is quite clear.

[CRoberson]

A derivative work under copyright law needs to contain creative expression from the original work. This is not the case when the original work is only used for adjusting weights. The legal situation is quite clear.

That makes derivatives of stockfish very clearly derivatives when the new "author" states something like "I changed 500 lines of code." Such programs have all the same data structures which I think is creative expression since data structures for chess can be created in many ways.