Machine translations are often brought up as a gotcha whenever I criticize LLMs. It's worth pointing out two things: Machine translations existed decades before LLMs, and yes, machine translations are useful. However: I would never in my life read a machine translated book. Understanding what a social media post is talking about in rough terms? Sure. Literature? Absolutely not. Hell, have you ever seen machine translated subtitles? It's absolute garbage.
If you have a fediverse account, you can quote this note from your own instance. Search https://mastodon.social/users/Gargron/statuses/116206874904164848 on your instance and quote it. (Note that quoting is not supported in Mastodon.)
AI Bullshit, mild Mastodon critique
RE: https://mastodon.social/@Gargron/116206874904164848
I always bring up auto-captioning as a (non-LLM) technology that is probably a net-win for accessibility, but that still led to Youtube ending up with *WORSE* closed captioning, because as soon as they got machine transcription, they dropped community transcription options.
Adding machine transcription to Youtube led to a local maximum where community options that often resulted in better transcriptions are now unavailable and everything is kind of meh.
RE: https://mastodon.social/@Gargron/116206874904164848
This is also absolutely true for software localization.
Translation tools are useful when I’m in charge of how I use them, and when I have contextual hints to help me interpret them.
But when a piece of software is localized exclusively using machine translations, not only are the resulting translations of poor quality (yes, even when done with LLMs — I’d even argue that they’ve gotten worse with LLMs), but also, as a user, I end up at a loss as to how to deal with them, except by switching the app back to English (if the option exists).
Suffice it to say that you’ve massively failed at your localization mission if the result is so bad that you’ve made me fall back to a language that I don't natively speak.