AI Contributions Discussion

[maryliag]

OpenTelemetry has been facing issues with low quality or large volumes of AI Contributions, which is creating more burden for maintainers.

This is a general issue for people to:

• provide feedback
• share what has been working and what is not
• provide ideas on how to deal with it
• volunteering to get more involved

[mx-psi]

I think this is one of the most important problems we face as a project and I'd be happy to work on solutions/improvements with other people.

Some notes on things discussed among some members of the Collector SIG last Monday:

• There was some support on making the AI policy more restrictive when it comes to documentation-related contributions and banning AI usage there, since human-written text feels quite important
• A wishlist:
  • CI checks that tell you if something is likely to be AI-written
  • Shared PR template with a "a human wrote this PR description" checkbox or other AI-related requirements and a workflow that fails if said checkbox is not marked
  • Some way to restrict issues that first-time contributors can work on
  • More automation on first-time contribution workflows (e.g. automatically remove assignees after a certain amount of time)
• AI contributions change the dynamics of contributions and promotion processes; contributors need to go further to demonstrate human involvement before getting promoted