← Back to context

Comment by mapmeld

3 days ago

I think it's an interesting perspective, because translation is one of the jobs that I (a) hear is the first to lose work due to AI, and (b) often used as an example of "acceptable" AI by people who are skeptics of LLMs and AI-generated art.

> often used as an example of "acceptable" AI by people who are skeptics of LLMs and AI-generated art.

As one of such people, I think there is a nuance to it. AI is great when you’re translating something to yourself. But when translating things for others, more caution and human judgement is needed. Espesially when translating instruction manuals, where bad wording could cause someone to injure themself.

  • Exactly, it's never about absolute results, it's always

    Expected Value (Upside, given time/cost savings + Downside, given %reliability).

    So, every task falls under a spectrum

  • This. I put things through Google translate all the time and they're always unreliable. Sometimes they're correct, sometimes I need to know roughly what the original said. Infamously, Google used to say "geiler Typ" meant "horny guy" when it means "awesome guy". Google used to think "geil" meant "horny" in general, which it can but not usually

  • Language is incredibly complex. I remember a TikTok from a bilingual English-Korean speaker comparing the English subtitles from a Squid Game scene to what was actually being said by the characters. The nuance and info density lost in translation made the subtitles feel completely remedial. Americans were basically watching a different show altogether.

There are translators and there are translators. Translating legal/business documents is a completely different thing from translating movies/books/games.

I can confidently say that LLMs do a better job than the average traditionally published fictions in my country, at least when the original works are in English. Every single time I watch a subbed movie there will be some lines noticeably wrong.

  • Yes, I've become very leery of artistic translation, in part because the paradigm of translators as adapters and localizers often ends up at odds with the job of faithfully and accurately representing the original material.

    The most egregious example I came across recently was where a friend enthused about some manga he was reading and I agreed to read a few chapters, only to discover that the translator has decided to render the countryside accents of western Japan (engaging with a protagonist visiting from Tokyo) by having them say 'y'all' and 'bless your heart' and other Southern USA tropes. I get the aspiration of the translator, but it was excruciatingly unpleasant to read. At that point, why not just say the protagonist was from New York and on vacation in Florida, or draw in some meshback caps on some of the characters and add alligators here and there in the background?

Translators already started losing jobs due to machine translation a decade ago (e.g. DeepL), before LLMs. Remuneration going down made it more difficult to make a living as a translator already then, even if you still received offers.

Well it's more than acceptable to translate e.g. web pages for reading, but it's not something you'd want to professionally publish.

Kinda conceptually similar to how typos and grammatical mistakes aren't a big deal if you're shooting off a quick text or email, but publishing if you've got typos in your advertising copy, in your resume, on your medicine label, etc. it's a real bad look.

Not all translations are the same. Literary translations are often works of art in and of themselves, and automating them would be missing the point entirely, like automating homework or weightlifting at the gym. I don't really know what's the state of the art, but I do buy that, on the other hand, translating toaster manuals or generic copy could soon be automatic.

  • Yup. If you are bilingual, you quickly realize how some translations are very bad. How some translations are very good. And how hard it is to translate. With dry, simple text, it might be easy. But when it involves art? Some jokes don't translate directly. There is pun. Sounds of words. Double meaning. Ambiguity. Cultural background. The creation of new words.

    It can be reasonably argued that some poetry can be impossible to translate from some languages to others. A poem might be explained, but by a lenghty, dissecting explanation, that completely loses the point of it.

    • Or if you compare a poetic translation to a literal one, of different translations of the same work to the same language to each other.

It'll be a similar theme for all facets of work involving any language, slowly - human or code. We'll parrot about humans in the loop this and that, but I think it'll be less humans in the loop over time and I think most people will even be willing to settle for a slightly more mediocre translation or coded project. It all comes back to our dopamine addiction, where we like fast feedback. And the oligarchs like tools to suppress wages. We will be our own demise for not advocating for either UBI or job protections, instead, happily using the technology while also rolling our eyes that it could never replace us.

I think the only people that genuinely think AI translation is viable are those that never read in the first place.

There are a number of games I've seen and played on Steam which use MTL to get the game playable in English and universally the translation is absolute shit. Sentences that don't transition into another, wordplay that becomes nonsense and a completely flat affect to everything.