- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
From the article:
…given the pretty massive and widespread architectural changes needed to make the implementation libified and memory safe, we decided that the codebase is not a derivative work that would require carrying forward the GPL license and have decided to release the code under the MIT instead.
I’m sorry but that pisses me right the hell off. I don’t care who it is, using this rewrite-by-LLM to work around the GPL is just going to be a pox on the global software landscape.
Basically the author is saying “if you can afford to spend $20K on tokens, buy your way around the GPL! Those with the cash can bend, nay break, the rules!”
This is just abhorrent to me. It’s STILL a derivative work, git already existed and these LLMs 100%, for sure, used knowledge from their training of the original C git implementation. Just because no one knows exactly how, where or when in the process it occurred, we know the LLMs all had access to original git in their training data.
rewrite by llm to workaround GPL
iirc someone started a company doing just that.
…and they deserve eternal damnation for the idea, IMO. Hrrmph. LLMs by definition can only make derivative works. That’s what they do, draw from their training on other works by humans. They are nothing without the training data. Data which was scraped from the entire f*cking Internet, without anyone’s permissions. Why is this so hard for devs to understand?
Yes, I am grumpy about this.