Comment by sbrother
6 days ago
I 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.
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.
You are adding your comments and translating them, thats fine.
If it was just a translation, then that adds no value.
You are joking right?
I mean we probably don't talk about someone not knowing english at all, that wouldn't make sense but i'm german and i probably write german.
I would often enough tell some LLM to clean up my writing (not on hn, sry i'm to lazy for hn)
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.
I think multi -language forums with AI translators is a cool idea.
You post in your own language, and the site builds a translation for everyone, but they can also see your original etc.
I think building it as a forum feature rather than a browser feature is maybe worth.
You know that this is the most hated feature of reddit ? (because the translations are shitty so maybe that can be improved)
6 replies →
That's Twitter currently, in a way. I've seen and had short conversations in which each person speaks their own language and trusts the other to use the built-in translation feature.
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
Better yet, I prefer to read some unusual word choices from someone who’s clearly put a lot of work into learning English than a robot.
Indeed, this sort of “writing with an accent” can illuminate interesting aspects of both English and the speakers’ native language that I find fascinating.
1 reply →
100%! I will always give the benefit of the doubt when I see odd syntax/grammar (and do my best to provide helpful correction if it's off-base to the extent that it muddies your point), but hit me with a wordy, em-dash battered pile of gobbledygook and you might as well be spitting in my face.
Yep, it’s a 2 way learning street - you can learn new things from non native speakers, and they can learn from you as well. Any kind of auto Translation removed this. (It’s still important to have for non fluent people though!)
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 honestly think that very few people here are completely non-conversant in English. For better or worse, it's the dominant language. Amost everyone who doesn't speak English natively learns it in school.
I'm fine with reading slightly incorrect English from a non-native speaker. I'd rather see that than an LLM interpretation.
...I'm not sure I agree. I sometimes have a lot of trouble understanding what non-English speakers are trying to say. I appreciate that they're doing their best, and as someone who can only speak English, I have the utmost respect anyone who knows multiple languages—but I just find it really hard.
Some AI translation is so good now that I do think it might be a better option. If they try to write in English and mess up, the information is just lost, there's nothing I can do to recover the real meaning.
> 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!
Google translate used to be the best, but it's essentially outdated technology now, surpassed by even small open-weight multilingual LLMs.
Caveat: The remaining thing to watch out for is that some LLMs are not -by default- prompted to translate accurately due to (indeed) hallucination and summarization tendencies.
* Check a given LLM with language-pairs you are familiar with before you commit to using one in situations you are less familiar with.
* always proof-read if you are at all able to!
Ultimately you should be responsible for your own posts.
I haven't had a reason to use Google Translate in years, so will ask: Have they opted to not use/roll out modern LLM translation capabilities in the Google Translate product?
1 reply →
You are aware that insofar as AI chat apps are "hallucinatory text generator(s)", then so is Google Translate, right?
(while AFAICT Google hasn't explicitly said so, it's almost certainly also powered by an autoregressive transformer model, just like ChatGPT)
> it's almost certainly also powered by an autoregressive transformer model, just like ChatGPT
The objective of that model, however, is quite different to that of an LLM.
I have seen Google Translate hallucinate exactly zero times over thousands of queries over the years. Meanwhile, LLMs emit garbage roughly 1/3 of the time, in my experience. Can you provide an example of Translate hallucinating something?
2 replies →
Google Translate hasn't moved to LLM-style translation yet, unfortunately
Hard disagree. Google Translate performance is abysmal when dealing with danish. In many cases its output is unusable. On the other hand, ChatGPT is excellent at it.
Google Translate doesn't hold a candle to LLMs at translating between even common languages.
IMO chatgpt is a much better translator. Especially if you’re using one of their normal models like 5.1. I’ve used it many times with an obscure and difficult slavic language that i’m fluent in for example, and chatgpt nailed it whereas google translate sounded less natural.
The big difference? I could easily prompt the LLM with “i’d like to translate the following into language X. For context this is a reply to their email on topic Y, and Z is a female.”
Doing even a tiny bit of prompting will easily get you better results than google translate. Some languages have words with multiple meanings and the context of the sentence/topic is crucial. So is gender in many languages! You can’t provide any hints like that to google translate, especially if you are starting with an un-gendered language like English.
I do still use google translate though. When my phone is offline, or translating very long text. LLM’s perform poorly with larger context windows.
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 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
> i don't think it is likely to catch on, though, outside of culturally multilingual environments
It can if the platform has built in translation with an appropriate disclosure! for instance on Twitter or Mastodon.
https://blog.thms.uk/2023/02/mastodon-translation-options
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.