AI rules

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:

1. AI applied to UseTheSource source code itself: no.
2. Experimental AI applications of {sofware,natural} language processing: yes, with caution.

AI applied to UseTheSource source code itself: no

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.

Experimental AI applications of {software,natural} language processing: yes, with caution.

All kinds of applications of LLM’s in UseTheSource projects have common issues. A UseTheSource contributor applying an LLM in their project must:

1. If code generation or transformation is the topic, see above for issues with licensing and other rights.
2. 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.
3. Manage and communicate LLM licenses to users (and their users).
4. Manage and communicate the environmental impact of the training and usage of LLMs.
5. Manage and communicate the effect on the level of expertise in the target community.
6. Give the user agency over the choice of LLM to use for each task, during download/install, or during webservice login.