I would like to raise a broader community question about fair participation for non-native English speakers in technical/policy discussions.
Today, I made a policy proposal on JuliaRegistries/General on Github that took me more a couple of hours to draft and which I had sharpened by AI to be as precise as possible for further discussion (the use of AI was disclosed).
I was told to “refrain from the use of AI in the discussion.”
While I understand the concern about low-effort AI content, there is imho an important difference between:
pasting unreviewed AI-generated output; and
using language or structure assistance to express one’s own argument more precisely in a non-native language.
A community cannot simply hold all of the following expectations at the same time:
participation should be international and inclusive;
English is the expected working language;
legal/technical arguments should be precise;
non-native speakers should not use language/structure assistance;
maintainers with native or near-native English and long institutional memory set the conversational standard.
That is not creating a fair playing field.
At the very least, I believe that in difficult cases non-native participants should be allowed to write the original argument in their native language and provide an English translation. The original-language version would then be the authoritative text, while the translation can serve as the shared version.
I am not promoting “AI slob” but I strongly believe non-native speakers should not be placed at a structural disadvantage when complex arguments must be made in English.
C’mon man, my complaint about AI slop was not about that post but a subsequent one. That’s the AI-iest of AI posts that felt very low-effort and didn’t even address what I wrote.
As far as I’m aware, nobody in the julia community has ever seriously suggested that nobody use translators.
Rather, what has happened repeatedly and to an obnoxious degree from certain users, is that these users ask an LLM to generate a response to someone, and then getting called out on it and pretending they just use it for translation, when nobody writes like that.
I am willing to share the long interaction I had with AI to shape what I thought to be a nice and precise policy draft.
I still believe that I am raising a legitimate issue and yes, it is not so clear cut, there is abuse. But making everything abuse, because there is abuse is wrong imho.
I still believe that there needs to be a more reflected consideration of AI use. While I see some nerving issues (too long winded, some hallucination). I see that it has improved tremendously and helps to sharpen a discussion – yes, especially if you are non-native.
Yes, absolutely, I’ve engaged with you through that long thread (100+ long posts deep) in very good faith and without complaint about translation. I did object to a post of yours this morning.
But then you come here and complain about my objection (and Dilum’s note), pointing to a post I didn’t object to and talking about the hours you spent drafting it when the post I complained about was a response written within ten minutes. It feels very disingenuous.
Yes, I use AI myself when I want to generate text and code, including in this very launch.
Here (and in that GitHub issue), I don’t need to generate text for text’s sake and I don’t need to read extra text just for the sake of it. I want to talk to you and hear your thoughts, not some bot’s overly wordy and smarmy chatter about it.
Here (and in that GitHub issue), I don’t need to generate text for text’s sake and I don’t need to read extra text just for the sake of it. I want to talk to you and hear your thoughts, not some bot’s overly wordy and smarmy chatter about it.
That “bot’s overly words and smarmy chatter” can be off-putting and I agree.
But you dismiss what I consider very prudent use of AI. All important posts are drafted by me first at least in the principle line of thought and are then polished by AI.
I am quite open to criticism but honestly for a discussion on Registry policy with legal and technical issues being subtle, I find NOT using AI to be more precise a bad use of time. I would not have managed to set up a reasonable Policy draft without the help of AI in reasonable time (time is money unfortunately).
I am not using AI here (mostly) and there are sure topics where we can talk personally and you can indeed hear my thoughts and voice, but for a policy discussion on GitHub?!?
Time is an important resource, I agree! And when one is asking for other people’s time and attention, requiring (primarily) human-written contributions is a useful filter.
If a policy is not important enough for you to devote the time to write the proposal yourself (+ AI for translation and minor edits), why should other people devote the time to read and debate it?
It is getting a tad unfair now: I put in a lot of effort to draft the proposal myself and in fact the term „outbound package license expression / OPLE“ was my invention. I was then accused of having „Claude“ hallucinate such terminology.
AI helped to make it became succinct. Maybe take a look yourself, before you assume things. Thank you.
I’m literally just responding to what you wrote: that you give the AI the “principle line of thought” but that “you would not have managed to set up a reasonable Policy draft without the help of AI”.
I’m not commenting on your specific policy proposal — we shouldn’t be duplicating that particular debate in this thread — but rather the question you raised in this thread on this forum, which is whether it is acceptable to post generative AI output in Julia forums beyond mere translation and minor edits.