End of Japanese community

It sounds like Mozilla just turned on the machine without consulting the human translators to see if the machine actually worked in a useful manner.

I have to say it feels like a really familiar, NGO-flavored disrespect, though: “we’re doing this favor for underrepresented language communities,” regardless of whether they want/need it or not.

“There’s only X number of you having to shoulder the load in XX sub-community, don’t you want us to impose a bunch of ‘help’?”

Well, no, if the choice is between a formidable volume of slop and a smaller but well-executed volume of volunteer labor-of-love…

(…I say as a person very much without all sides of the story, and shooting from the hip a bit. I don’t mean to impugn anybody’s intentions, and I imagine at the end of the day we’re all on the same side here.)

Sure we had machine translation before, but it was still a little off. Now the latest language models get us 99.9% there, so they are judged good enough to deploy at scale. What results is a weird twilight zone where everything is in your language, except it feels kind of wrong and doesn’t really communicate in ways specific to the culture from which the language is.

You’re in France, you search for something, a thread pops up with everyone interacting in French - seems reasonable enough, but it just reads kind of weird? Then a message is entirely out of place, and you realize that you’re reading an English language thread translated to French.

Or your mom sends you a screenshot of a Facebook thread in her native language that has her worked up - and reading it, you realize it’s an LLM translation of something that should have no bearing on her.

Same with various support pages on websites - it all reads mostly fine until you hit a weird sentence where the LLM messed up and then you’re transported back to the reality that what you’re reading was not authored by anyone who can actually operate in that language/culture.

There’s a lot of nuance in language beyond the words - how you express disagreement in English is not how you express disagreement in Japanese, how you address the reader in French is not the same as in Korean, etc. Machine translation flattens all modes of expression into a weird culturally en-US biased soup (because that’s where the companies are headquartered and where the language models are trained).

I have no illusions that this trend will reverse - high quality translation work is skill and time consuming, and thanks to LLMs anyone on Earth can now localize anything they want in any language they want for ~free in ~0 time.

The weirdest part is seeing this bubble up to the real world. I’ve been hearing young people use turns of phrases/expressions that I recognize as distinctly American, except not in English.

The classic linguist response to this, which I subscribe to, is “no language is fixed, language is ever evolving in response to various external cultural pressures“. Which is true. But it doesn’t make our post-LLM language landscape any less weird.