Add AI use policy to README

[MonkeyDo]

Following discussion in our weekly meeting, adding a proposal for an AI use policy.

We might want to copy these guidelines in CONTRIBUTING.md and PR templates in each project as well because I expect that contributors opening AI-slop PRs are not going to be reading the guidelines carefully...

[mayhem]

Thanks for taking the first stab!

[mayhem]

Given our adopted acceptable AI use policy "OK, to use, but disclose" I feel that discourage is a bit too strong here. We could soften it, or we could strengthen it for first time contributors:

"We forbid the use of AI tools in communications and code contributions for first time contributors".

[reosarevok]

I think discourage is just fine, personally - but this could be more specific if we want. I think we should for sure:

• Discourage it in communications for everyone
• Discourage it for first time contributors, because we want to see what they can do without AI
• Allow it for code contributions later down the line, as long as the rest of the policy is followed

[mwiencek]

I'm not sure I'd personally hold a first-time contributor to a different standard unless they're planning to apply to GSoC, where AI use should be forbidden outright.

If the first-time contributor is already an experienced developer and knows how to "properly" use AI as part of their workflow...well, I don't particularly want to become a judge of what they can do without AI, as long as it's disclosed and their changes are "personally understood, tested and reviewed."

I'd also hold ourselves to the same standard: can we discourage others when we already have our own significant and succesful use of AI running in production? Of course Zas "personally understood, tested and reviewed" this project, but at the same time, the first sentence of this document can be read to cast disapproval on it.

[Aerozol]

Nice! The main thing is that the policy is solid, and it looks good to me. In terms of writing I think it's longer than it needs to be and only one or two things should be bolded, if really necessary. Shorter = more people will read.

I would suggest something like (up to you if you want to integrate any of these edits):

AI use policy

We discourage the use of AI tools in communication and code.

Undisclosed use of AI or Large Language Models, or inability to explain AI-generated contributions, will lead to your PR being closed without further review.

You may use LLMs as part of your workflow - with explicit disclosure - if you understand the issue and any problem solving has been done by you. This wonderful little open-source bubble has a humanistic philosophy, and prioritizes community, critical thinking and robust and long-lasting code over productivity and speed.

[MonkeyDo]

OK, I think we have arrived at a happy medium.
Can I get a thumbs up to merge?

We can always tweak it if we get feedback or questions from contributors.

AI use policy

We don't allow blindly using AI tools in code contributions or communications, and discourage it in most cases.
If you do use LLMs as part of your workflow, you must explicitly disclose it.

Do not submit a PR you haven't personally tested; inability to explain your code is grounds for immediate rejection of the PR.
We want to see you (not an LLM) show understanding of the issue and do all the problem solving.

This is the core of our work as developers. Open-source prioritizes community, critical thinking and robust and long-lasting code over productivity and speed.

[mwiencek]

It sounds very reasonable to me. Thanks for tackling this @MonkeyDo!

[MonkeyDo]

Thank you for the input and feedback everyone!