Hello everyone, I am Niranjan from India. I am a free software hacker. I am wondering how LLM-generated code should be treated in free software world. In 2024, when the discussions around this topic started, mainly because of GitHub Copilot, they were mainly centered around how LLMs are inherently developed with unethical methods. Now, in the last two years, LLMs have become even more omnipresent. Initially, I also carried the stigma towards them and didn’t even touch any of the popular LLMs, but lately, because of my dayjob, I had to use Claude. While using it, I got tempted to use it for many free software ideas that I had stalled only because I didn’t have the required coding skills in those domains. My skills are mostly in LaTeX programming and I write original free software in that community, but other than that, I had many ideas in my TODO list that almost seemed unachievable because of the constraints I had as a person. Claude very easily brought the ideas in reality, though. It was pleasant to see how the ideas that were stalled for years got materialised within few hours. I was wondering what to do with such projects, though. Because as a free software enthusiast, I am very much willing to distribute them to the public, but I want to have an ethically and philosophically sound position while doing that.
I found a term called ‘transvestment’ [1] while reading around this and my current position is somewhat like that. I want to transfer something that is built using unethical systems to a system that I consider more ethical. The unethically built AI is a reality. I or people’s collectives would most probably not be able to overthrow trillion dollar industries in the current scenario. If I develop and distribute the so called ‘vibe-coded’ software, I, at least, help my communities in some way instead of staying idle and never materialising them. We all know that LLMs were trained using the knowledge of thousands of free software hackers, but by abstaining to engage with LLMs because of this, I am not adding anything to the communities either. I either can live a ‘pure’ life myself that adheres to my free software ideals or I can make some philosophical compromises and make my communities stronger. I am tempted to choose the latter. In fact, I have to choose the latter many times because of practical constraints, e.g., donating to FSF from India is, at the moment, only possible with the non-free crypto-currency, because none of the other alternatives work whenever I try. I read the AI policy of GCC [2]. If I understand correctly, the copyright status of LLM-generated code is itself dubious in most countries. Since contributions to a GPL-licensed project essentially need copyrights, non-copyrightable or legally dubious code cannot be accepted/merged. I find it a very well thought and pragmatic move, but my next question is, suppose I want to release a free software on Savannah, can I use CC0 license to disclaim/waive the copyright (if applicable) and affirm that no action would be taken if the copyright is deemed inapplicable. That helps my free software activism, creates some utility that my community may find helpful and still be ethically sound by not claiming any rights on the things I didn’t work on. I am posting this here as I want to know how Savannah would treat vibe-code submissions. If acceptable, should they go to Savannah GNU or Savannah non-GNU? If not acceptable, what would you all recommend? I don’t mind making them public domain too, to be honest, but I think CC0 is better and solves the potential legal problems. Best, निरंजन The world needs so much change that the capitalists start clarifying, “free as in free beer and not as in freedom”. [1]: primer.commonstransition.org/archives/glossary/transvestment/ [2]: gcc.gnu.org/ai-policy.html
signature.asc
Description: OpenPGP digital signature
