| title | AI rules | ||||
|---|---|---|---|---|---|
| keywords |
|
The application of AI -in particular Large Language Models (LLMs)- is interesting for UseTheSource projects. Although most of the practical consequences are already outlined elsewhere, we explain some of the consequences here for our contributors. Two different aspects:
- AI applied to UseTheSource source code itself: no.
- Experimental AI applications of {sofware,natural} language processing: yes, with caution.
When AI is used to generate or transform source code of UseTheSource projects be aware that you might fail to be compliant with the Contributor License Agreement:
- Each contribution which I submit is and will be an original work of authorship and I can legally contribute it to open-source.
- To the best of my knowledge each contribution I submit will not violate any copyright laws, trademarks, patents, or other intellectual property rights.
Conclusion: AI can not be used to literally generate or transform code or (inline) documentation for our projects.
All kinds of applications of LLM's in UseTheSource projects have common issues. A UseTheSource contributor applying an LLM in their project must:
- If code generation or transformation is the topic, see above for issues with licensing and other rights.
- If publication of AI generated language is a feature (arxiv, conference, issue tracker, online discussion), then see above, with code replaces by natural language text.
- Manage and communicate LLM licenses to users (and their users).
- Manage and communicate the environmental impact of the training and usage of LLMs.
- Manage and communicate the effect on the level of expertise in the target community.
- Give the user agency over the choice of LLM to use for each task, during download/install, or during webservice login.