policies: Add AI Policy.

[jcolp]

This AI policy serves as a beginning and may evolve over time. It is a modified version of what was proposed on the Apache Arrow project and discussed on their own mailing list:

apache/arrow#48952

This fundamentally does not forbid the use of LLMs but does require many things of the individual using them in order to be accepted, and explicitly states that we may outright close such PRs.

[maximilianfridrich]

Very good. The note

Watch for AI's tendency to generate overly verbose comments, unnecessary test cases,..

could potentially be made even more general. In my experience the generated code itself can also be "too verbose", e.g. repeated/redundant code, dead variables/code and this verbosity increases the burden of the reviewers/maintainers. Just a comment, it's good as it is.

[jcolp]

Very good. The note

Watch for AI's tendency to generate overly verbose comments, unnecessary test cases,..

could potentially be made even more general. In my experience the generated code itself can also be "too verbose", e.g. repeated/redundant code, dead variables/code and this verbosity increases the burden of the reviewers/maintainers. Just a comment, it's good as it is.

Good thought - I've tweaked it some.

[phoneben]

Boundaries = freedom.
Thanks, jcolp , for always being direct and clear.