site banner

Friday Fun Thread for January 10, 2025

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

1
Jump in the discussion.

No email address required.

Obnoxious issues that arise when you read Chinese Cultivation Novels:

At some point, when the chapter count is well past 4 digits and/or when there's a hiatus, human translators (usually unpaid fans) can drop out, and you run into MTL (machine translation).

This might not sound like a problem today, after all, LLMs are pretty good at the job. To test that myself, I've converted chapters from (translated) English to Chinese and back, and find near perfect fidelity.

No, dear reader, the issue is that the MTL can be old. And it's non-trivial to find the original Chinese sources.*

Cue me weeping in agony as the best translation of a favorite novel changes character gender and names every few paragraphs, and translates what was previously "Land of Sin" as "Naughty Land".

Some frankly insane bastards persevere nonetheless, becoming one with the Dao of MTL, and self-reportedly no longer see the broken Mandarin Matrix but grokk the underlying intent. Unfortunately, often at the cost of being unable to process normal English.

Fuck it, I'm going to do dig down the Chinese version and throw it into Google Gemini, a million token-wide context window can't hurt.

*Most places where you can read Xianxia in English use unauthorized or outright pirated translations. And these sites all steal from one another, so if a bad translation becomes the default, good luck finding a better one.

Some frankly insane bastards persevere nonetheless, becoming one with the Dao of MTL, and self-reportedly no longer see the broken Mandarin Matrix but grokk the underlying intent. Unfortunately, often at the cost of being unable to process normal English.

Betcha Claude can grok the underlying intent and create a less-borked translation too, and any damage to its sanity would be isolated to only that chat. Care to provide a sample?

I've actually spent about 4 hours yesterday using Gemini 1206 to translate as I go, the 2 million token CW meant I can throw in dozens of chapters and have it produce perfectly pleasant text. There are a few stylistic quirks, or perhaps they were the original translator's quirks that aren't present anymore. It certainly doesn't garble gender and do anything too funky.

I'm sure Claude would be great at the task, but even the original GPT-4 would consider it trivial.

Here's the site I've been grabbing the raw text from if you want to take a look: https://www.piaotia.com/html/7/7095/6107291.html

The horrid "official" translation can be found on:

https://novelfull.com/forty-millenniums-of-cultivation/chapter-2771.html

https://novelfull.com/forty-millenniums-of-cultivation/chapter-2771.html

Wow you weren't kidding about that translation quality. And yeah probably any recent LLM can do it, but that 2M context limit is pretty sweet when you need it.