@thunderbird
"This feature would be fantastic in Thunderbird, don't you think?"
NO!
There are HEAPS of features which are MUCH more important then this. GOOD mail-encryption for example.
Translation is also no core-element of an email-client.
So whoever needs such should make a addon out of it!
@thunderbird yes, it would be!
But together with possibility to quickly look at the original if the auto translation is vague, as e.g. google translate does.
@thunderbird not a bit, no. For one, I suspect it will be based on some #LLM crap, so thanx but no thanks.
And as others pointed out, TB is a MAIL client. If I get emails in a language I don't know, odds are, its #Spam anyway.
@fedithom@thunderbird if you ever tried to start working in a country which speaks other language than yours, you wouldn't be so sure. In EU, it is normal.
I'm not sure what you're saying. If I move to another country, and I don't know the language yet, I try to learn it. I STILL wouldn't want my emails translated by some LLM junk, either way ;)
@fedithom@thunderbird yes, obviously, one should learn the language to better communicate, but it takes time, during which you still can be contacted by people who don't speak your language or just by broadcast emails to all employees, the majority of which would be locals.
Surely, it's up to you if you personally want to use such feature, but it can be useful and it is nice to have it.
@nanoelquant@thunderbird not if it's LLM based it isn't :)
And since #Mozilla has already happily announced LLM crap, I don't see a reason to assume otherwise in this instance :)
Firefox Translations uses machine learning for translation, but it deploys all the necessary tools locally on your machine, and it's pretty lightweight. No data is transmitted out. If you don't trust it, audit it -- it's open source (https://github.com/mozilla/firefox-translations).
This is a huge privacy advantage to using something like Google Translate.
MANY people work in global organizations, or receive support emails from people who don't speak their native language.
First off: Massive thank you for even replying. That's far from the norm.
I read that, yeah. What I can't find is info on training data. With what data was Firefox Translations (or Marian underneath it, I guess) trained? Was that obtained in full acknowledgement of copyright / usage rights, and with awareness of people who may have trained it?
I know it's easy to lump all LLM usage in IT in with ChatGPT and it's brokenness, but a good bit is deserved, I'd say.
Add comment