Comment by Palmik
2 days ago
Nothing wrong with a GPL-like viral license for the AI era.
Training on my code / media / other data? No worries, just make sure the weights and other derived artifacts are released under similarly permissive license.
2 days ago
Nothing wrong with a GPL-like viral license for the AI era.
Training on my code / media / other data? No worries, just make sure the weights and other derived artifacts are released under similarly permissive license.
Well, I would say it should be like that already & no new license is needed. Basically if a LLM was ever based on GPL code, its output should be also GPL licensed. As simple as that.
Licenses like GPL are built on top of an enforcement mechanism like copyright. Without an enforced legal framework preventing usage unless a license is agreed to, a license is just a polite request.
We need countries to start legally enforce that. Nothing will change otherwise. I stopped open sourcing my code and LLMs are one of the big reason.
Wouldn't you want the code generated by those models be released under those permissive licenses as well? Is that what you mean by other derived artifacts?
That's how I interpreted it at least
It really should be like that indeed. Where is RMS? Is he working on GPLv4?
If model training is determined to be fair use under US copyright law—either legislated by Congress or interpreted by Federal courts—then no license text can remove the right to use source code that way.
> then no license text can remove the right to use source code that way.
At least in the US.
Quite what happens if another country ordered, say chatGPT, to be released under the AGPL since it was trained on AGPL code, who knows.
You can follow him on https://stallman.org/ What is he doing? I believe still giving talks and taking stance on current day political issues. Additionally I believe the last few years where quite turbulent so I assume he is taking life at his own pace.
RMS is probably greatly behind the technical news at this point. I mean, he's surfing the web via a email summary of some websites. Even if he doesn't condone of how the internet is evolving, he can't really keep up with technology if he doesn't "mingle".
He's also 72, we can't expect him to save everyone. We need new generations of FOSS tech leaders.
I am gen-z and I am part of the foss community (I think) and one of the issues about new generations of FOSS tech leaders is that even if one tries to do so.
Something about Richard stallman really is out of this world where he made people care about Open source in the first place.
I genuinely don't know how people can relicate it. I had even tried and gone through such phase once but the comments weren't really helpful back then on hackernews
https://news.ycombinator.com/item?id=45558430 (Ask HN: Why are most people not interested in FOSS/OSS and can we change that)
10 replies →
Interesting. Is there a license that acts this already ?
That is a complete fools errand. If it ever passes it would just mean the death of Open Source AI models. All the big companies would just continue to collect whatever data they like, license it if necessary or pay the fine if illegal (see Antropic paying $1.5 billion for books). While every Open Source model would be starved for training data within its self enforced rules and easy to be shut down if ever a incorrectly licenses bit slips into the models.
The only way forward is the abolishment of copyright.
I don't follow. If the model was open-sourced under this GPL-like license (or a compatible license), then it would follow the GPL-like license. If the model was closed, it would violate the license. In other words, it would not affect open-source models at all.
Similarly, I could imagine carving out an exception when training on copyrighted material without licence, as long as the resulting model is open-sourced.
> If the model was closed, it would violate the license.
Training is fair use. The closed models wouldn't be impacted. Even if we assume laws gets changed and lawsuits happened, they just get settled and the closed source models would progress as usual (see Bartz v. Anthropic).
Meanwhile if somebody wants to go all "GPL AI" and only train their models on GPL compatible code, they'd just be restricting themselves. The amount of code they can train on shrinks drastically, the model quality ends up being garbage and nothing was won.
Further, assuming laws got changed, those models would now be incredible easy to attack, since any slip up in the training means the models need to be scraped. Unlike the big companies with their closed models, Open Source efforts do not have the money to license data nor the billions needed to settle lawsuits. It would mean the end of open models.