Re: [PATCH v4] CONTRIBUTING.d/ai: Add guidelines banning AI for contributing

["G. Branden Robinson" <g.branden.robinson@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 5118 bytes --]

Hi Alex,

At 2025-10-15T20:11:10+0200, Alejandro Colomar wrote:
> I prefer that they use badly written English by a human, than
> good-looking English written by an AI.

That's a tolerable and defensible position, but it is not the element of
your policy that is drawing pushback from Carlos, Sam, or me.

(...assuming I'm understanding Carlos and Sam correctly--as always I can
speak only for myself.)

> The principle I'm following is: consider an AI as a fancy version of
> a chat with Jia Tan.
[snip]
> Are you using the AI to walk?  Then it can't possibly affect the
> quality of the contribution.  Are you using it to translate something?
[snip]
> > I think Carlos is drawing a line in a good place.  By grounding
> > policy on the process of _production_ of content, rather than the
> > process of knowledge _reception_ by contributors (all of whom are
> > imperfect) we better avoid the pitfalls of hallucination in both
> > natural and artificial intellgences.
> 
> Let's consider again the case that AI is a fancy version of a chat
> with Jia Tan.  Should we trust contributions where Jia Tan has
> influenced in any way?  I strongly believe that we shouldn't.

I don't think the Jia Tan scenario is a useful litmus test.

Jia Tan was presumably a state actor, maybe FSB/SVR, DIA/NSA, or MSS.

Another possibility is that Jia Tan was the agent of a huge metanational
corporation,[0] like those characterized by the erstwhile FAANG
acronyms.

Yet another is that Jia Tan was a front for a highly motivated
individual or small group, possibly a criminal enterprise in the
ransomware business.

All of these possible identities have in common a protection from civil
and possibly even criminal penalties for breach of contract or
fraudulent attestation of any warranty or guarantee or ask of them.
Criminal enterprises with sufficient resources routinely buy impunity
for their crimes and tortious offenses against the public.[1][2][3][4]

Nothing that "Signed-off-by:" implies is binding upon Jia Tan.  Not in
any practically enforceable sense.

The Jia Tans of the world can and will lie to all of us, and the Linux
man-pages project will be wholly unable to hold them to account.

> I think of the three concerns (legal, quality, ethical), the first one
> affects code produced directly by the tool, but quality concerns apply
> as well to code influenced by the tool.  And obviously, the ethical
> concerns apply to *any* use of AI.

Richard Stallman, too, attempts to persuade software developers of the
world to adopt his ethics.[5]  To what extent do you think he succeeds?

That's not a rhetorical question; projects that employ copyleft licenses
foreclose contributions from people who hold that BSD-style licenses, or
dedication to the public domain, are "the only true free[dom]".[6]

I don't suggest that you shouldn't hold the views that you do, or even
that you shouldn't express them in material that Linux man-pages
contributors are likely to see.  I _am_ saying that there seem to be
some accessibility applications of "AI" that are beneficial to some
potential contributors (and maybe some already existing), specifically
in the case of machine translation of existing English in the Linux
man-pages _to_ a person's native language so that they can better
understand the system documented, and subsequently contribute revisions
and additions to the project's pages, in English, from the metaphorical
sweat only of their own brow (meaning: without "AI assistance"), that
may be inexpertly composed but that other contributors like you and me
can wordsmith to a satisfactory level.

In a game-theoretic sense, when you devise a social contract, you should
do so in anticipation that your counterparties are those you can expect
to earnestly, if not necessarily perfectly, abide by it.  They are
"cooperators".  A social contract cannot constrain the actions of
"defectors" who abide by the contract, if at all, only to build up
reputation for the day they betray the community to their advantage.[7]

Even more broadly, this is the insight that informs Wilhoit's Law.[8]

Regards,
Branden

[0] https://en.wikipedia.org/wiki/Mars_trilogy
[1] https://www.reuters.com/markets/us/us-authorities-set-unveil-settlement-with-binance-source-2023-11-21/
[2] https://www.justice.gov/usao-edny/pr/goldman-sachs-resolves-foreign-bribery-case-and-agrees-pay-over-29-billion
[3] https://subscriber.politicopro.com/article/2024/06/billionaire-michael-saylor-to-pay-40m-over-tax-fraud-charges-00161273
[4] https://www.reuters.com/sustainability/boards-policy-regulation/meta-investors-zuckerberg-reach-settlement-end-8-billion-trial-over-facebook-2025-07-17/
[5] https://www.fsf.org/about/what-is-free-software
[6] https://manpages.debian.org/unstable/mailcap/run-mailcap.1.en.html

[7] The title is cringe, but the piece is useful nonetheless.
    https://tryingtruly.substack.com/p/how-generous-tit-for-tat-wins-at-life

[8] https://crookedtimber.org/2018/03/21/liberals-against-progressives/#comment-729288

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]