Advances (not really) in machine translation

February 16, 2023

So, I translate for a living, and just like a lot of people these days, I also to some degree worry that one day my job is going to go poof.👻 But even though the thought crosses my mind from time to time, I’m pretty sure it’s never going to happen, at least not with my language pair of Japanese to English, and it’s obvious why.

Context is everything

Machine translation, while now outputting more natural sounding sentences, has the exact same problem it’s had from the very start: it doesn’t pick up on nuance, sarcasm, certain slang, and all that. Especially with a language like Japanese. And frankly, I don’t see how this can be fixed.

日本語上手ですね

Still, machine translation engines are tools that I use at my job every day.🙋‍♀️ I don’t think MTEs are bad, they’re actually really helpful! But I’m well aware of their shared shortcomings—there’s no getting around those limitations.

Deepl can be pretty accurate, but it will still give me the exact opposite of what the Japanese source text says on occasion.😬 Sometimes the source text will contain errors, or a translation will need to be done while checking reference material along the way. At the very least, there will always need to be a human to look over and edit machine translated text.

And now time for a related rant…👴📢☁

We just recovered from having to hear about Crytpo and NFTs all the time, and now here comes ChatGPT.😭 (Insert multiple disclaimers on the complexity of this topic and how I am but a regular degular person who sounds like a boomer. Also, not all deep learning models.) Girl, the hype I’m seeing seems a bit excessive.

“ChatGPT simply makes it too easy for users to generate responses and flood the site with answers that seem correct at first glance but are often wrong on close examination.”

“Ask the bot to write a biography of a public figure, for example, and it may well insert incorrect biographical data with complete confidence.”

The Verge – AI-generated answers temporarily banned on coding Q&A site Stack Overflow

“Something something this is just the beginning,” yes, but I still have my doubts. “Something something, people said the same about Wikipedia,” noted. The thing that annoys me isn’t that it “can’t think like a human” or whatever, but that it’s being talked about as if it can or that it soon will.

ChatGPT does seem like it could be useful for generating work emails, poems, short stories, marketing copy, various fluff text.👀 It’s like how machine translation is most useful for stuff like privacy policies, instruction manuals, and other such documents with stiff, unimaginative writing. But those aforementioned usage cases for ChatGPT, for your average person, don’t match the extraordinary claims I’ve been hearing. If all the hype is for its “future potential,” then to me that’s just more wishful thinking. I mean, the potential is going to be that companies won’t have to pay as many people. Much progress, many innovation.

I’m just being a bit negative here because this is something that is going to be of no great benefit to you or me. It all seems very vague atm and nothing’s regulated. We’ll just have to see what happens, as the train has already left the station.

4 comments

Older Posts