6 Comments
User's avatar
Mefistofele's avatar

This was a very interesting read, but I only have a minor point to discuss. At the beginning of the article, AI is quoted as excelling in translation, but this clashes with everything I've been doing as both a hobbyist translator for friends, as a majoring student in translation studies and as a subtitle translator.

At the very best, AI can be a somewhat useful tool for professional users to lighten the load, but fully autonomous translations are always flawed, incoherent or downright hallucinated.

Park Jee-hoon's avatar

That’s interesting, perhaps there’s a difference in terms of method? From what I’ve seen, translations by LLMs like Copilot, Grok, etc. are quite biased, but built-in browser “webpage translation” tools seem to work decently well. I do agree that fully-AI translations nearly always have flaws, though (losing nuance, rephrasing whole sentences, hallucinations as you said, etc.).

Mefistofele's avatar

When the translation isn't very long, the output is serviceable, usually, that's true, and the dedicated translators from browsers work better than more general LLMs. I'm not really sure though on the reason for that, I'm not an expert on AI programming

Eman Mohammed's avatar

Thank you for this. It is the terminal contradiction of the bourgeoisie, inshallah automating their own collapse.

giacomo catanzaro's avatar

it's not only technically dubious, it's fundamentally unsound and impossible to have a machine be able to always know what is relevant in a given situation https://arxiv.org/abs/2506.10077

Dante Douglas's avatar

Great article. I am a huge, *huge* AI skeptic and I have to admit, one of the things that worries me the most about the modern PRC is their wholehearted embrace of genAI. I worry that they, just like the west, are going to get hit with the inevitable backlash of that tech (both in the sense of environmental damage but also the breakdown of any services that rely on genAI).