My boundaries are absolutely only about me. Using spell check is one thing, but if you outright can't write without using an LLM prompt then no, I don't want to read it thinking a person wrote it. If that doesn't catch on, I'd sooner move to a whitelist approach or stop reading altogether than be forced to read it.

You seriously don’t want to read content that was hand-written in another language and translated by LLM? That seems extremely parochial.

I am seeing this on the OpenStreetMap forums, which are an international affair, and it really annoys me. We get well-meaning mappers who join a thread in a language not their own (in case something is discussed within a national community) using LLM-translated posts.

For Dutch, this is extremely annoying¹. It's not that you can't translate to and from Dutch, it's that you will not pick up the nuances in the text written by people with a decent proficiency in Dutch (like the way written and spoken Dutch can be really rather direct, which can translate to quite impolite English, and really improper German), and technical and domain-specific content (e.g., traffic regulations) gets butchered.

I much rather see someone responding to a Dutch thread do so in English if they can't write Dutch, because then at least I can see if the translation from Dutch is going wrong somewhere, instead of having to figure out why that person isn't making sense by going through two passes of an LLM… Been there, done that. Besides, if I'm replying I can do so in English too, and avoid having LLMs mangle my words.

So yes, I too abhor having to deal with any form of communication where an LLM sits between the other person and myself. I find it exceedingly rude.

1: For other languages too, but as a native Dutch speaker this one is easy for me to see.

I said "LLM prompt" because I meant just that, a prompt. I described my stance on it more thoroughly here: https://news.ycombinator.com/item?id=42829068

I absolutely do not want to read that. I want google to stop sending me that. Either it’s written in French or English and I can read it directly, or it’s written in another language and I can ask for automatic translation myself, but do not lie to me about who wrote it and in what language.

I’m so tired of translation slop. I live in France, and when I search for building related stuff in French I have to wade through pages of translation slop to find stuff written with the actual building standards and codes in mind. Avoiding sales pitch, AI, and translation slops is getting really tiring when you’re looking for contextualized expert knowledge.

I don't want to read anything generated by an LLM, now or ever.

I am trilingual. Sometimes, Google would auto-translate their docs into the local language, despite my browser and account language being set to English. I hate this. Monolingual people may not fully grasp how much languages differ in the exact details of how you write; a translation will always alter the text and when done without a human mostly rewriting the entire thing by hand, it would make it more confusing, meandering, and unpleasant.

This nicely sums up my distaste for the recent Lex / Zelenskyy interview. I feel like the auto-translation was a mistake, and I would have preferred anything else.

If non-native speakers (including myself, fwiw) want to post grammatically correct comments, there's a fairly straightforward solution: learn grammar and use a spell/grammar checker. Have the courage to write your own words and the decency to spare the rest of us from slop.

People who depend on LLMs to polish their words will run into the same problem as people who rely on autocomplete functionality: their language skills will suffer.

There's nothing wrong with using tools to check written text, but I'm wary of blindly accepting any suggested fixes. If I see a red underline I'll consider whether the word is actually fine first (English is not a static language, and spelling dictionaries are not complete), and if it looks wrong I'll try fixing it myself before reaching for the suggested fix.

Then either you edit the results as suggested in TFA or those comments are in fact not yours. Grammatically correct or otherwise.

Comments that have been translated are yours or at least widely regarded as such. An aversion to AI isn't going to change that.

In the US at least, translators own the copyright of their translation. That is to recognize the complexity of translating meaning and not just words from one language to another.

Sure, but if you ask almost anyone who wrote a work of fiction or whatever that was translated, they mention the author, the translator often not even coming into the picture at all. Ultimately, most people don't really care about translators, complex job or not.

Definitely. I'm not saying it's solely the work of the interpreter (clearly not), but it is a significant intellectual contribution. I do not think this contribution has remotely been made obsolete by artificial translation.

I tentatively agree - if the core idea buried within the text is unique enough then I'm not sure I care how much the text has been laundered. But that's a big IF.