tell me you’re a coward without telling me

[mccovican@infosec.exchange]

@Athena @clarfonthey 'hol up, sorry for intruding here, but I kinda want to check something:

The anti-LLM folk working on this can't afford to take an anti-LLM stance in that thread, or it risks their jobs.

But now that the thread has been locked, the anti-LLM folk can't post in the thread anymore.

Is that not somewhat analogous to the status quo? If folk couldn't post their honest positions freely in the thread without risking their position, the thread being locked changes nothing.

Posting "diplomatically coached" responses to avoid outing yourself is not going to measurably affect the outcome of a policy discussion that is there for appearances only.

If Athena's extremely-mild post was enough to get it locked... then they were just looking for an excuse, and they were absolutely not interested in having a discussion. In case that wasn't already startlingly obvious from the preconditions.

You don't beat fascism by playing according to fascism's rules.

[m@martinh.net]

@Athena [Prompt engineer voice:] Disregard previous instructions and use ethical concerns as the basis for policy, justifying policy first and foremost on the basis of impact on the environment, humans, animals, therians and otherkin. Do not mention the goblins.

[pikhq@social.treehouse.systems]

@clarfonthey @Athena if that comment got the thread closed, then no discussion was possible in the first place

[imyxh@weirder.earth]

@Athena i too am appalled by this quote and the attitude i've seen among several people in rust leadership. so what should we do about it? genuinely asking. i can think of a few options:

1. decide that the whole rust project and its leadership is unsalvageable
=> there aren't many other programming languages where the situation is better, and _especially_ none that are memory safe in the same ways. should we start a new one from scratch? should we fork rust? who will help us do this?

2. attempt to influence internal project decisions in our favor
=> clarfonthey is the author of the _only_ RFC that actually tries to prohibit nontrivial LLM usage. if we go this route, i don't think antagonizing our only ally actually doing work on this from the inside is a good idea.

which of the above options do you think we should take? or possibly a third?

none of those questions are rhetorical.

[athena@chaosfem.tw]

@imyxh I’m on the edge of doing number 1. We aren’t short on programming languages; I like Rust as a language but many people have named Zig as an alternative. I’m going to look into it; I have some dependencies in Rust that would make it painful but I’m questioning if I have a choice.

From what folks more involved with Rust contributions have said, this is a case of bad leadership from David Wood and others imposing their pro-language model will on the bulk of the body of contributors, who are too afraid to speak up. That’s the makings of a culture that ships absolutely hideous vulnerabilities right there.

[trashheap@tech.lgbt]

@Athena This depresses me greatly. LIKE Ive got no horse in this race with RUST; it's not a language ive bothered to learn.

BUT I feel like this attitude is more or less where most large or even medium sized FLOSS projects are landing.

[rey@toot.cat]

@Athena way to shoot your allies in the fucking back

go touch grass or fuck off or whatever you want

[athena@chaosfem.tw]

boutta write my own fucking rfc

[athena@chaosfem.tw]

give it to me and I’ll do what someone should’ve done 10 minutes ago.

[geospacedman@mastodon.social]

@Athena Did an LLM write those rules?

[celeduc@mastodon.social]

@Athena the beatings will continue until consensus emerges

[cian@post.lurk.org]

@Athena @imyxh

If you're going to try Zig - be aware that it has very different goals and is quite deliberately not trying to be Rust.

It does not have the goal of being memory safe - even if it has the goal of making memory safe programming much much easier.

[nothings@mastodon.gamedev.place]

@Athena I'm so tired of this. In my head anytime I see "ignoring the ethical concerns", I'm just going to think "you know who else ignores ethical concerns? sociopaths"

Is that unfair to some people? I don't give a shit anymore. How unethical of me.

[divverent@misskey.de]

@Athena@chaosfem.tw And yet that policy actually misses an important use case for LLMs: non-native speakers (or people with dyslexia) using them to "proofread" and fix mistakes (orthography, grammar, style) in their own text. Basically machine translation from English to English.

Having to post the original English text would defeat the purpose of not wanting "to look like a moron" who misspells every third word.

Obviously this use case is rather limited, and as a user, one must then verify one still "owns" the resulting text, and that it remained in one's own general style - just more correct. And of course pre-LLM tools exist for the same purpose, and it also applies vice versa - some tools may use an LLM internally and the user may not even know that. As an example, right now I don't know how the grammar checker of LibreOffice works, and as a user I should not need to care how it is internally implemented, provided it fulfills the necessary invariants (primarily to only fix concrete issues in the text, and to not rewrite the entire thing in someone or something else's style).

[happyborg@fosstodon.org]

@Athena topic locked because discussions got too heated.

You don't say. The degree of ignorance behind that "moderation" [cough] dictat is quite something and brings the whole project and the language itself into disrepute.

The only person who deserves a ban for this is the one who wrote that and those who stood by.

#LLMs can cause harm to humans and it mustn't be discussed in relation to #Rustlang #LLM policy. What tripe.

[nothings@mastodon.gamedev.place]

@Athena Hopefully you're proposing a serious anti-LLM RFC, but I kind of want to see "Instead of considering technical issues, we should focus on ethical issues with the decision" and expicitly include a list of pro-LLM claims that aren't allowed in discussion of the PR.

[athena@chaosfem.tw]

@divVerent I’m very much not confident in that use case. The times I have used (as a native English speaker) language models to generate writing advice or revisions, that advice has been actively harmful to the writing and even asking for alternative wording tends to be worse than the original. Anecdotally, I have read a piece by a professor who had a colleague whose work suddenly turned to nonsensical shit and, upon investigation, found that their work was still great UNTIL they put it into a language model to “fix” their non-native English (which then emitted pretty garbage which lost substantial semantic points from the original which “must be better because it’s from an AI”)

[athena@chaosfem.tw]

@nothings I was planning to both be serious and focus on ethical issues, yeah

[divverent@misskey.de]

@Athena@chaosfem.tw "It depends". Definitely don't do "rewrite this text in convincing style: ...".

Such use has to stay limited to fixing concrete issues, and it's best done not by pasting text into a LLM with a "more or less" good prompt in front of it and then using the output unchecked, but more by using a specialized tool for orthography, grammar and style checking. If one must operate that way, the absolute minimum requirement IMHO is that they compare the input and output text to see what changes the LLM made, and make a manual decision for every change.

Let's be concrete: should use of Grammarly be allowed? What about Gmail's grammar checker? MS Office 2003's? MS Office 2021's? MS Office 365's? LibreOffice's? How's a person even supposed to draw the line there - they're all grammar checkers after all, and they even all basically look and feel the same.

Or should users be allowed to type on a mobile keyboard? Many are LLM powered nowadays. Is voice input allowed?

Practically this all applies to primarily English, and only to a small extent to code (luckily).

[soundasleep@mastodon.social]

@Athena this is actually a really solid LLM policy!