docs: add AI policy and link from CONTRIBUTING.md - #16583
Conversation
rmuir
added
the
skip-changelog
|
|
||
| **AI should not be used to generate comments when communicating with | ||
| maintainers**. We expect comments on our projects to be written by humans. We | ||
| may hide any comments that we believe are AI generated. |
There was a problem hiding this comment.
can we add a note like "It is OK to use AI to translate text authored by a human writing in a differnt language." Or maybe just move the last paragraph immediately after this one?
There was a problem hiding this comment.
I don't know what you are asking. can you re-read it again? The current document is very specific about exactly how to do this.
I don't want to read slop.
There was a problem hiding this comment.
I am fine with the last paragraph about machine translations. Adding the original language in the text and only add translation as blockquote, clearly marked, is fine. People may translate with AI, but they should instruct their AI agent to not also add slop. "Keep sentence structure and stay away from addingtext or extra formatting" is a good AI prompt.
in general, Google Translate is much better as it keeps original sentence structure, so I'd prefer to ask people to use a conventional translator instead of a chatbot.
There was a problem hiding this comment.
I was suggesting we could move the last paragraph up so it is the second paragraph. That way it comes right aftrer the part that (in my opinion) it is modifying or clarifying, so the reader isn't held in a suspended state until the end, but you know it's fine as it is,, just trying to make it a little clearer
There was a problem hiding this comment.
@msokolov If you want to wordsmith it, do you mind making a followup PR?
I was just imagining such a policy could be heavily controversial and was attempting the route of "use an existing one verbatim that appears to already be working".
Does not mean it is perfect for our needs, just a way to take a step in the right direction.
There was a problem hiding this comment.
yes, please go ahead. We can fight the grammar battles and such in another thread, LOL
uschindler
left a comment
uschindler
left a comment
There was a problem hiding this comment.
Looks good!
Maybe some additional comments as it fits the same thing: Please also add a text somewhere to forbid intensive use of Emojis in PR descriptions. I know those are mostly AI generated, but we should maybe still add this to the pull request template, too. I have seen PRs where every second line contains an emoji or colored bullet points, or checkmarks in wild colors everywhere. Not all of them are AI generated (I am still unsure), but people should also not create those colorful descriptions! This has to end....
An emoji in the normal discussion of the PR is fine, but not a colorful issue/PR descriptions (and pages of AI slop - of course).
|
I also added a PR with a SECURITY.md file for our repo: #16585 Feel free to review it, too. The mass of rejected security issues which could have also done as a normal PR were too immense, so I'd like to clarify that tampered index files are no security issue .... and so on. |
+1 on this! I have also seen AI generating overly verbose code comments, unnecessary test cases and benchmarks, maybe we can add a note around that as well later. |
|
+1 I really like this proposal, thanks for taking the time to write it! |
|
For things like "don't make unnecessary emoji, tests, comments", all that kinda stuff probably belongs in AGENTS.md IMO, not here. This file is for human contributor. I'm already concerned it has a near-zero chance of actually ever getting merged, so we should try to keep the concerns separate to simplify the situation. Just my opinion |
That's all fine. I just wanted to add my findings when reviewing PRs. Actually some emojis are also coming from persons (not sure which culture encourages that, maybe also because of AI suggest and they copy AI's behaviour). As said, some Emojis were in PR descriptions which were heavily human written (because of typos), but still had those empojis. Therefor I suggested to add this somewhere. But yes, an AGENTS.md would be fine, too. Best would be an AGENTS.md saying "stay away and f*ck yourself". I think nobody has something against this PR, lets get in in after maybe 72 hours of waiting time. I will do the same for my SECURITY.md file (see other PR you already reviewed). Uwe |
|
@uschindler Yeah I use them sometimes. But I try to not do it excessively. Sometimes you need a 💩 or a 🤯. Sorry, could not resist. 🍻 |
6 participants
Per discussion on dev, add simple AI policy and link it from CONTRIBUTING.md.
Policy is sourced from astral's policy
PR template today already links to CONTRIBUTING.md:
Issue templates already link to CONTRIBUTING.md (via github's default widgets):
So the content is reachable from the key places where it counts. If there is trouble, we could always do more aggressive stuff as a followup (e.g. direct links to AI policy from these templates, required checkboxes on those templates that you've read it, etc etc). But it seems best to avoid that kind of thing unless it becomes strictly necessary.