Histomat of F/OSS: We should reclaim LLMs, not reject them
"The materialist response isn't to reject the new technology. It's to evolve our licenses to encompass it."
This is very close to how I feel:
“I want my code to be used for LLM training. What I don't want is for that training to produce proprietary models that become the exclusive property of AI corporations. The problem isn't the technology or even the training process itself. The problem is the enclosure of the commons, the privatization of collective knowledge, the one-way flow of value from the many to the few.
[…] The question is: who owns the models? Who benefits from the commons that trained them? If millions of F/OSS developers contributed their code to the public domain, should the resulting models be proprietary?”
The problem with LLMs, in my mind, isn’t that they exist to begin with. They’re genuinely useful. It’s an issue of power dynamics and consent. Training material shouldn’t be taken from creators without their knowledge or participation; the resulting models should not be owned by a small number of wealthy corporations instead of a collectively by the people who put their work and creativity into them.
Rather than pretending that LLMs will go away and we’ll revert to a pre-AI state of being, what would it look like for these things to be built and run equitably and sustainably?
I’m inspired by ETH’s Apertus model — a fully open, multilingual LLM that is built with shared openness and representation in mind. More will likely follow (and I’m beginning to hear similar stories in specific fields like journalism). Can that be part of our future? Possibly, and the first step is not necessarily rejecting AI but rejecting how it’s run today. Software, as we’ve learned, is best built for the people, and open has historically won eventually. Open, consensual models will almost certainly win too.
[Link]
