CODA now has a released page.

[Ras]

In which case this little no-money, fun, hobby/retired computer chess thing is becoming nasty, unpleasant and potentially dangerous.

I don't think so. Respect the licences, and you're good. The problem isn't with chess programming, it's that the AI companies didn't think about legal aspects at all. They just went ahead like it's the Wild West. They created tools that do put its users at legal risk. So they will have to work on traceability. That's a work that is already being done anyway in that also for non-programming use, indicating the sources of the information is making progress.

Otherwise, as soon as the legal departments in tech companies figure out that Claude etc might accidentally pull them into the GPL zone without documenting that, they will simply forbid Claude use. Hence, Anthropic (and others) must address this topic, or else their business case will collapse.

[chrisw]

In which case this little no-money, fun, hobby/retired computer chess thing is becoming nasty, unpleasant and potentially dangerous.

I don't think so. Respect the licences, and you're good. The problem isn't with chess programming, it's that the AI companies didn't think about legal aspects at all. They just went ahead like it's the Wild West. They created tools that do put its users at legal risk. So they will have to work on traceability. That's a work that is already being done anyway in that also for non-programming use, indicating the sources of the information is making progress.

Otherwise, as soon as the legal departments in tech companies figure out that Claude etc might accidentally pull them into the GPL zone without documenting that, they will simply forbid Claude use. Hence, Anthropic (and others) must address this topic, or else their business case will collapse.

Rybka respected the licences and look what happened there. There are nasty shitty people in computer chess and they just love ganging up and wrecking people and other engines, especially the successful ones. Obviously, use of Claude, which will be ubiquitous, is going to allow the hunters plenty of opportunity. Not a healthy place to be, so why be in it?

[AndrewGrant]

Asked Claude a little more and it appears it already knows these conversations and that it is itself involved! You could regard the below as the response of Claude:

Yeah so I am not going to read that.

You can't stop the future Andy, it has been tried before and it did not work, better find a good compromise, as it happened before.

I don't think "the future" is me having to read AI generated content in a forum for humans. I suspect the conclusion you and Chris draw is that I am 100% anti AI, use non of the tools, and condone all use? Interesting estimation of my position if so, but not even remotely close.

If our most veteran members can't see the loss we'll suffer if humans delegate their interactions with others entirely to the AI, then we really are lost. If anything, you guys should hold my position. And I should be the young-naive one arguing for "progress".

[AndrewGrant]

AG has basically dropped the case with respect to the thirty or so GPL engines ...

To be clear, I have not "dropped the case". I am arguing the AGPLv3 case, which is a clear and evident winner by simple human or AI observation, as affirmed by Ras, and the author of Coda himself -- as it is the most effective entry point.

Slapping a GPLv3 label on your project does not magically grant you access to include all other GPLv3 projects in it. Perhaps one won't agree that -all- LLM output is an active violation of the collective IP of the world --- but explicitly pointing your LLM at someone else's IP and saying "What can we use from this", certainly is. And, as Ras mentioned, all those engines are not GPLv3. I'm sure there is a mixture. I know of at least some MIT engines in there. Presumably some other licenses as well.

---

As a personal matter, I do care about honouring such obligations. Torch for example can link against Pyrrhic ( a fork of an MIT licensed Fathom, of an MIT licensed Syzygy probe tool ). However, that code is not linked against in Torch distributions on the website, as there is no good way for us the convey to the user directly what is being done. I have been trying to get chess.com to open a /open-source/ page, which will explicitly list out all such libraries that make it to end users.

Furthermore, the following blurb appears at the end of any email from me, to someone who has purchased Ethereal.

Additionally, you should know that Ethereal is licensed under the GNU GPLv3, just like many other open source chess engines. The GPLv3 is a permissive license, that grants you, the user, the right to request a copy of the source code, the right to redistribute that source code, and the right to build your own projects using the source code so long as those projects comply with the GNU GPLv3.

Ethereal, and all of the tools I use to build, train, and test Ethereal, are open source. They can be found on Github, at https://github.com/AndyGrant. These two paragraphs don't concern most people -- but I think it is important to include here. Computer chess would not be in the incredible state it is in today if not for the willingness of developers to release their code under the GPLv3, and the willingness of users and other developers to respect the license.

[mclane]

So what is AI allowed to do ?


In 1990 Ed Schröder wrote one of the strongest 8 Bit dedicated chess computer Modules called Mephisto MM5.

There have been emulators running Eds MM5 6502 engine with high speeds.

But now an AI Project reengineered the assembler Code of the Eprom in C and made an UCI engine out of it.

Ed Schroeders Engines were the strongest Dedicated 8 bit 6502 chess computers.
Neither spracklens nor kittinger or any other guy was capable to build stronger 8 bit engines.

The 22,000 lines of C were compiled as UCI engine running in almost any PC.
Depending on the PC you have MM5 suddenly running 12.000 mhz fast on your PC.

In 1990 MM5 ran on 5 mhz speed.

So is this legal ??

Is AI allowed to translate assembler code, copyrights to ed schroeder, into C source code and run it as uci engine ??

AI is attacking or challenging us day by day.

Ed said: you can’t stop the future.


IMO people like Chris and Ed are pioneers.

They made computerchess in the early days, when many of us were not even born or focussing on computerchess.

I had the great chance that i began computer chess even before Ed and chris were working.
I began 1978.
But as a customer. A very young customer.


What is allowed .
What is possible for the AI.

I am worried. I wonder for how long the job programmer makes any sense if the AI replaces programmers.

[jasper.sinclair]

What is allowed .
What is possible for the AI.

I am worried. I wonder for how long the job programmer makes any sense if the AI replaces programmers.

At this point in time, AI is helpless if left on it's own. In all disciplines, it requires an expert human to guide it, to prompt it, to manage the architecture, goal, and details, of the project.

There's no doubt that CODA required an enormous effort and time from Adam, to guide Claude (and work with it as a team)...meticulously implementing, managing, and testing any suggestion it may have. It's called engineering! and is critical to the success of the project.

Maybe this will change in coming years...but or now, the idea of an AI acting independently, without human guidance, is a fantasy.

So as it stands right now, AI is improving human lives in every aspect, rapidly improving the world by accelerating scientific research, automating tedious workflows, and tackling global challenges like disease detection, food security, and climate change.

You need not worry about AI taking over the world, as Hollywood might suggest...those films are geared toward naive teenagers.

Again I defer to Chris, the only one who seems to be making any sense!

All that Claude has done is to massively speed up the development cycle, it’s not doing anything new.
Cycle is, and everybody does it, scan GitHub etc for new source releases. Go through the code looking for what’s changed/ideas your own program lacked. Select the most promising (Elo) idea, code it, tune the weights (if any), test if better, keep idea or junk it. Repeat cycle.
Nothing new, nothing unlawful, just faster, more efficient and fully automatable.

[AndrewGrant]

Again I defer to Chris, the only one who seems to be making any sense!

All that Claude has done is to massively speed up the development cycle, it’s not doing anything new.
Cycle is, and everybody does it, scan GitHub etc for new source releases. Go through the code looking for what’s changed/ideas your own program lacked. Select the most promising (Elo) idea, code it, tune the weights (if any), test if better, keep idea or junk it. Repeat cycle.
Nothing new, nothing unlawful, just faster, more efficient and fully automatable.

There are separate and distinct conversations happening in here. I am not arguing that AI cannot speed things up; I am not arguing that AI is not inevitable; I am not arguing that it is not a helpful tool for developers. If you've read those claims into my posts -- then I suggest going back and reading them fresh, with the clarifying point.

I am merely pointing out that these LLMs are ruthlessly violating copyright all over the place.

I've demonstrated that, at least ChatGPT, is perfectly capable of providing a user with GPLv3 covered code, verbatim, without informing them. And we've demonstrated, that in the case of Coda, Claude has taken verbatim code from some of the reference engines, including Viridithas, and as such has to follow the license attached to Viridithas. A detail which Adam was not actively informed about by Claude, as far as we can tell.

!! The above paragraph contains zero opinion statements. It is not really up for the debate !!

Any time you have Claude write code, it is a lottery as to whether the code coming out is already a protected property by someone else. It should be no surprise that this is the case, when the prompts are to explicitly explore an engine's code. But such a situation can arise even without that. You might prompt Claude to implement some Search feature, and find yourself with a cut-and-paste of some existing work.

[jasper.sinclair]

Again I defer to Chris, the only one who seems to be making any sense!

All that Claude has done is to massively speed up the development cycle, it’s not doing anything new.
Cycle is, and everybody does it, scan GitHub etc for new source releases. Go through the code looking for what’s changed/ideas your own program lacked. Select the most promising (Elo) idea, code it, tune the weights (if any), test if better, keep idea or junk it. Repeat cycle.
Nothing new, nothing unlawful, just faster, more efficient and fully automatable.

There are separate and distinct conversations happening in here. I am not arguing that AI cannot speed things up; I am not arguing that AI is not inevitable; I am not arguing that it is not a helpful tool for developers. If you've read those claims into my posts -- then I suggest going back and reading them fresh, with the clarifying point.

I am merely pointing out that these LLMs are ruthlessly violating copyright all over the place.

I've demonstrated that, at least ChatGPT, is perfectly capable of providing a user with GPLv3 covered code, verbatim, without informing them. And we've demonstrated, that in the case of Coda, Claude has taken verbatim code from some of the reference engines, including Viridithas, and as such has to follow the license attached to Viridithas. A detail which Adam was not actively informed about by Claude, as far as we can tell.

!! The above paragraph contains zero opinion statements. It is not really up for the debate !!

Any time you have Claude write code, it is a lottery as to whether the code coming out is already a protected property by someone else. It should be no surprise that this is the case, when the prompts are to explicitly explore an engine's code. But such a situation can arise even without that. You might prompt Claude to implement some Search feature, and find yourself with a cut-and-paste of some existing work.

Yes I agree!
and God forbid anything else from reading or referencing Ethereal (splendously repleat with Stockfish ideas & code), or another top 30 engine!
What a complete and utter travesty of justice this Coda is!

[AndrewGrant]

Yes I agree!
and God forbid anything else from reading or referencing Ethereal (splendously repleat with Stockfish ideas & code), or another top 30 engine!
What a complete and utter travesty of justice this Coda is!

Did you miss the posts from the Coda author, where he acknowledged there are issues with the licensing?

If your position is that intellectual property does not exist, or that LLMs are free to digest whatever they want, or that open-source is just one big bin indifferent to the licenses attached, then that would be a coherent argument.

I will summarize again the most straight-forward argument, since it seems some posters are having a hard time following the conversation.

- Coda recently changed the license to GPLv3.
- Coda contains code from Reckless and Viridithas.
- Reckless and Viridithas are licensed under the AGPLv3.
- The GPLv3 is not compatible with the AGPLv3, as it rescinds some obligations.
- By publishing AGPLv3'ed code, marked as GPLv3 code, Coda is attempting to re-license Reckless/Viridithas.

[AndrewGrant]

Yes I agree!
and God forbid anything else from reading or referencing Ethereal (splendously repleat with Stockfish ideas & code), or another top 30 engine!
What a complete and utter travesty of justice this Coda is!

Details matter here, so we're going to clarify the post you made from ChatGPT, which incorrectly stated the license and origins of the Pyrrhic library, which was your attack point for Ethereal. You posted the following from ChatGPT:

Ethereal's syzygy.c is not an original implementation—it wraps Pyrrhic, which is itself a fork of Fathom, Ronald de Man's GPL Syzygy probing library that originated from the Stockfish ecosystem. Ethereal explicitly includes and interfaces with Pyrrhic rather than reimplementing the probing logic from scratch.

Ethereal's syzygy.c is an original implementation, but it is just a wrapper. Pyrrhic code is what is doing the work. Pyrrhic is a fork of Jon Dart's fork of Fathom. Fathom is licensed under an MIT license. The original version of Fathom, from Basil Falcinelli, is also under an MIT license. The original Fathom has the following copyright notice:

(C) 2013-2015 Ronald de Man (original code) (C) 2015 basil (new modifications)
Ronald de Man's original code can be "redistributed and/or modified without restrictions".
The new modifications are released under the permissive MIT License:

If you go to Ronald's probetool repo found here (https://github.com/syzygy1/probetool), you will see "The license for this code is: do whatever you want with it (but don't hold me liable for any games your loses)."

So ChatGPT has mistakenly determined that Fathom, and its fork Pyrrhic, are GPLv3 code. They are not GPLv3. They are MIT code. Ethereal is distributed with a copy of the MIT license associated with Pyrrhic/Fathom, dating back six years or more, found here: https://github.com/AndyGrant/Ethereal/b ... ic/LICENSE

Ethereal fully honours the license obligations associated with the Syzygy probing code. IF we determine that these tools should have been under the GPLv3, then it would be proper for Fathom and Pyrrhic to update their licenses; and for upstream users like Ethereal to adjust their license. Then, things like Torch would be in a tough position as Torch cannot use GPLv3 code, so it would be a problem to distribute Torch with Syzygy support.

Hopefully of all the posts in this thread, at least this one gets digested by a human. As it is quite clear. And anyone is readily able to google these projects and look at the license themselves. It is quite surprising that ChatGPT got it wrong.