The FSF considers large language models
6 months ago
- #Copyright
- #LLM
- #Free Software
- The FSF is conducting a survey to understand free-software projects' stances on LLM-authored code.
- No current plans for GPLv4; FSF is considering adjustments to the Free Software Definition first.
- LLM-generated code raises concerns due to non-free models and training software.
- Key questions include copyrightability of LLM code and potential infringement risks.
- Suggestions for projects accepting LLM code include metadata collection and clear marking.
- Blurred lines between LLMs and assistive technology pose challenges for outright bans.
- Distinguishing LLM-generated from human-authored code is difficult, requiring human accountability.
- Developers may lack awareness of risks associated with LLM-generated code submissions.
- LLMs cannot cite inspirations, making it hard to trace code origins.
- Session concluded with raised awareness but no definitive answers on LLM-related concerns.