The Linux Kernel Organization now lets developers submit AI-generated code, as long as it complies with the guidelines, licensing, and attribution requirements

https://www.xda-developers.com/linux-kernel-now-allows-ai-written-code/

73 points · 15 comments · view on lemmy.world

15 Comments

onlinepersona@programming.dev · 39 pts · 160d (3 replies)

AI agents MUST NOT add Signed-off-by tags. Only humans can legally certify the Developer Certificate of Origin (DCO). The human submitter is responsible for:

  • Reviewing all AI-generated code
  • Ensuring compliance with licensing requirements
  • Adding their own Signed-off-by tag to certify the DCO
  • Taking full responsibility for the contribution

That's fair. Nobody has to know you wrote it with a not, that's impossible to detect, but you have to own it and be ready for the discussions that follow.

peterhorvath@mastodon.de · 1 pts · 159d (2 replies)

@onlinepersona @Innerworld And what do you think, will the AI agent be ready to the following discussion?

towerful@programming.dev · 8 pts · 159d (1 reply)

Well, no. But then it gets rejected, and further PRs that also fail the check will likely get you banned from contributing.

The human is responsible.
If the code or PR fails, the human has to own that.
If the human fails to own that, the human gets banned

peterhorvath@mastodon.de · -1 pts · 159d

@towerful Also you know, it is only matter of time and it will.

heliotrope@retrofed.com · 20 pts · 160d (8 replies)

Rule-wise, this seems fair.

Regardless, if AI usage continues to increase in this manner, I'll likely be driving NetBSD, AROS, and FreeDOS by the end of the decade.

Maybe even a little TempleOS or ZealOS, for flavour.

Dumhuvud@programming.dev · 6 pts · 160d (6 replies)
magikmw@piefed.social · 3 pts · 160d (5 replies)

It's just regarding labeling. It's unenforcable to have a project "clean" of AI.

Dumhuvud@programming.dev · 5 pts · 160d (4 replies)
magikmw@piefed.social · 4 pts · 159d (3 replies)

Ok I see the intent of BDFL is different, but the linked document only mentions labeling - I can only assume the low quality etc. issues are handled as a judgement call, and in that way I consider the "No AI whatsoever" rule unenforceable.

If I use an LLM to generate code under my suprvision, review, quality check and test to be up to standard, how would it be detected I used AI if I don't label it so? They'll look for em-dashes in comments?

soc@programming.dev · 3 pts · 159d (2 replies)

"Let's not have rules, because some may break them!"

🤡

magikmw@piefed.social · 0 pts · 158d (1 reply)

Rules without enforcement are just self-deception.

soc@programming.dev · 1 pts · 157d

Then keep deceiving yourself. 🤷

Dumhuvud@programming.dev · 1 pts · 160d
[ removed ]
misk@piefed.social · 5 pts · 160d

Given that nobody is able to guarantee that code used for training was used according to it’s license, this means no hallucinated code in Linux. Nice.

g_blob@programming.dev · 4 pts · 159d

It will comply but will not compile

abcdqfr@lemmy.world · 2 pts · 160d

It really has come a long way in a relatively very short time in terms of quality and, well, shitting under the rug