alt Hacker NewsI strongly agree with this sentiment and I feel the same way.
The one exception for me though is when non-native English speakers want to participate in an English language discussion. LLMs produce by far the most natural sounding translations nowadays, but they imbue that "AI style" onto their output. I'm not sure what the solution here is because it's great for non-native speakers to be able to participate, but I find myself discarding any POV that was obviously expressed with AI.
Replies
When I occasionally use MTL into a language I'm not fluent in, I say so. This makes the reader aware that there may be errors unknown to me that make the writing diverge from my intent.
Non-native English speaker here:
Just use a spell checker and that's it, you don't need LLMs to translate for you if your target is learning the language
Agreed, but if someone uses LLMs to help them write in English, that's very different from the "I asked $AI, and it said" pattern.
> I'm not sure what the solution here
The solution is to use a translator rather than a hallucinatory text generator. Google Translate is exceptionally good at maintaining naturalness when you put a multi-sentence/multi-paragraph block through it -- if you're fluent in another language, try it out!
I have found that prompting "translate my text to English, do not change anything else" works fine.
However, now I prefer to write directly in English and consider whatever grammar/ortographic error I have as part of my writing style. I hate having to rewrite the LLM output to add myself again into the text.
As AIs get good enough, dealing with someone struggling with English will begin to feel like a breath of fresh air.
I think even when this is used they should include "(translated by llm)" for transparency. When you use a intermediate layer there is always bias.
I've written blog articles using HTML and asked llms to change certain html structure and it ALSO tried to change wording.
If a user doesn't speak a language well, they won't know whether their meanings were altered.
one solution that appeals to me (and which i have myself used in online spaces where i don't speak the language) is to write in a language you can speak and let people translate it themselves however they wish
i don't think it is likely to catch on, though, outside of culturally multilingual environments
Maybe they should say "AI used for translation only". And maybe us English speakers who don't care what AI "thinks" should still be tolerant of it for translations.
I wrote about this recently. You need to prompt better if you don't want AI to flatten your original tone into corporate speak:
https://jampauchoa.substack.com/p/writing-with-ai-without-th...
TL;DR: Ask for a line edit, "Line edit this Slack message / HN comment." It goes beyond fixing grammar (because it improves flow) without killing your meaning or adding AI-isms.
If I want to participate in a conversation in a language I don't understand I use machine translation. I include a disclaimer that I've used machine translation & hope that gets translated. I also include the input to the machine translator, so that if someone who understands both languages happens to read it they might notice any problems.