Single post
jump to replies
Came across this #GoLang Dev mailing list post about #AI tool usage, and it seems a bit tighter(?) than the #Linux Kernel. In the kernel, they want attribution of the model and other metadata as a Co-Author. But in #Go, it seems they have a don’t ask, don’t tell policy. If an AI tool is used, they do not want any Co-Author line and will ask for it to be removed before merging.
Effectively, both projects are fine with AI tools being used, but attribution seems to be taken differently.
https://groups.google.com/g/golang-dev/c/4Li4Ovd_ehE/m/8L9s_jq4BAAJ
3 visible replies; 1 more reply hidden or not public
back to top
@bryan Just a clarifying question... Do you mean 'tighter' in the sense that the Go project 'disallows' AI contributions by making invidiual contributors solely responsible for the code they submit? Whether those contributors then used AI or not is kinda beside the point?
@cuboci yeah that’s probably bad word choice from me.
They both clearly accept patches from AI Tooling just attribution and submission details are different.
#Go accepts #AI assisted work but they have tight rules that forbid mentioning a Co-Author/Assisted that isn’t a person. The person submitting the patch, AI enhanced or not, is the sole author. You basically don’t say and if you do they tell you to remove a byline in the commit first.
In the Linux Kernel it seems they prefer to see an “Assisted-By”with information about the model used. See this document:
https://docs.kernel.org/process/coding-assistants.html
@bryan This begs the question how both projects get around the problem of non-copyrightable AI code, as Russ Cox said in that thread. Whether the co-authorship of AI is made explicit or not shouldn't matter, I reckon.