consensus on LLM usage (was Re: Thoughts about LLM usage)

Holger Levsen <[email protected]> Tue, 11 Aug 2026 19:54:07 +0000

Newsgroups	gmane.linux.debian.devel.vote
Message-ID	<[email protected]>

On Mon, Aug 10, 2026 at 09:13:25PM +0200, Matthias Urlichs wrote:
> I tend to agree, except for one point I feel fairly strongly about:
> "consensus".
> 
> A consensus is, by definition, a result which most (if not all) can live
> with. Not just some majority, not even some supermajority.
> 
> Speaking of supermajorities. Assume that Proposal A "ugh-no-AIs-ever" wins
> by 8:3 above Proposal B "allow-AIs" (which itself narrowly wins by 100:99
> above Further Discussion), but gets tossed out because it doesn't attain the
> required 3:1 supermajority (it does attempt to change the Social Contract
> …). If that happens, the result – "allow-AIs" wins – is not something I'd
> call something the "project as a whole" feels, much less a consensus.
> 
> (I know that this is highly unlikely, esp. given the existence of Proposals
> C, D, E and F …)
> 
> The discussion so far was very interesting and did inform my approach to the
> LLM question, but it definitely didn't result in any sort of consensus IMHO.

agreed, i'm very very sure there won't be any consensus about LLM usage
no matter how we'll vote.


-- 
cheers,
	Holger

 ⢀⣴⠾⠻⢶⣦⠀
 ⣾⠁⢠⠒⠀⣿⡁  holger@(debian|reproducible-builds|layer-acht).org
 ⢿⡄⠘⠷⠚⠋⠀  OpenPGP: B8BF54137B09D35CF026FE9D 091AB856069AAA1C
 ⠈⠳⣄

Which Large Language Model do you prefer? The one that claims trans people
don’t exist or the one that doesn’t want to talk about Tian'anmen Square
in 1989? (@[email protected])

signature.asc (application/pgp-signature, 833 B)