Histomat of F/OSS: We should reclaim LLMs, not reject them (writings.hongminhee.org)

A recent blog post by a free and open source software (F/OSS) advocate highlights growing concerns about how AI companies exploit open source code without acknowledging individual developers' contributions or providing adequate legal protections. While the author of the post suggests withdrawing from platforms like GitHub to avoid exploitation, another voice in the community argues for a more constructive approach: reclaiming the technology. This perspective emphasizes the need to evolve licensing to support the use of F/OSS for training large language models (LLMs) while ensuring that the resulting models remain accessible to the community. The proposed solution involves creating a "training copyleft" license, which would allow the use of F/OSS in LLM training but require that any models trained on such code must be released under compatible licenses. This approach aims to maintain the core principle of F/OSS—that collective knowledge remains communal rather than privatized by corporations. The discussion underscores a crucial turning point for the AI/ML community, emphasizing that rather than rejecting LLMs, developers should focus on shaping the governance of these technologies to ensure equitable distribution of benefits derived from the commons. The conversation reflects the ongoing struggle between technological advancement and the protection of open source ideals.