By appearance it seems that certain people in the kernel are trying to force through AI policy without community engagement.

This is not a good look.

It may not reflect the reality of the situation (though it's hard to think otherwise at the moment), but sadly in this area - optics matter.

It matters because the press are going to report this as 'linux kernel accepts AI patches'. It's inevitable unless we are very careful about how we pursue this.

And that kind of reporting will inevitably lead to a sea of AI slop.

I did raise all these points on-list, it doesn't really feel like much of a community when they are essentially, ultimately, ignored.

@ljs My main concern is that the person that proposed the policy is clearly not doing it from an objective point of view just by looking to the language they used.

Which makes me wonder.. why? Are you trying to score at your work? In addition, I am concerned that several people added their +1 without mentioning this..

Anyway, I am not a maintainer so it doesn't matter but that doesn't look good. And I agree with you, the press is going to report this as you mentioned.

@ljs That proposal clearly at least misunderstands kernel development as "a task solving a computational problem in a vacuum" :-) It is hard to interpret that text by any other means.

@ffmancera yes I think a little more sensitivity on how this was done at the very least should have been applied.

Doesn't help that the whole thing that kicked off the discussion came from Sasha sending a quiet series that iterated quickly with the same concerns (adding llm tooling dotfiles).

A lot of the issues were explicitly raised in the thread also.

I replied. Some of the doc is great, others I have concerns with.

https://lore.kernel.org/all/11eaf7fa-27d0-4a57-abf0-5f24c918966c@lucifer.local/

@ljs personally, when i see patch think "why that person needs this" or 'why company X needs this" and stuff like that. Bots don't need any of the changes they supply, which is a problem (at least for me). I don't want to serve the needs of those who don't have needs because well, it's fucking crazy for instance :-)

My reply to the thread - https://lore.kernel.org/all/11eaf7fa-27d0-4a57-abf0-5f24c918966c@lucifer.local/ - FWIW.

Some parts of doc are great. But really needs to be made more firm in my opinion.

And community needs to be invovled

@ljs Yeah that is nice. I hope they take it into account. Thanks for sending this :)

@ffmancera glad at least one person thinks my opinion is of value :)

@ljs @ffmancera sir i suspect sometimes even like two people secretly think your opinion is of value, perhaps more

@lkundrak @ffmancera sometimes it feels like this is not the case :) but thanks.

Ape brain misbehaving atm though

@ljs that's what happens in corporate projects and i fear it's going to go the same path as Fedora did, i.e. "shut up nay-sayers and let's go AI!"

@tragivictoria I mean it's not even anti-AI so much as 'let's be sensible about AI' remaining _neutral_ about AI basically.

So nobody can really complain on either side of the discussion...

The problem here is that we really need community buy-in and to be cautious re: impact of 'hey kernel wants AI' type press.

Luckily the doc posted is at least mostly decent.

The concern is the 'a committee privately decided to do this' is not the way including a very loud advocate who also works for nvidia but doesn't ever mention that...

You could argue he ought to recuse himself but at the same time wouldn't we all in tech... :)