@Gargron I'm pretty sure J.R.R. Tolkien would view LLM's as an abomination

@Gargron I had what I thought was a really great concept for a book game, but I'm not a talented artist.

I think it was 21 or 22 I started using AI to try and make the images I could see in my head.

Must have run over 1000 iterations with various prompts, negative prompts, and many hours of research. I got exactly zero usable artwork.

It was equally unhelpful with descriptive filler text even when presented info on a silver platter. AI = Garbage Out.

@Gargron Has everyone already heard of the brilliant machine translation for the saying "Out of sight, out of mind"?

"Invisible, idiot."

@Gargron 💯💯💯💯💯💯‼️

@Gargron
Hello
How are you?

@Gargron all your base are belong to us, man...

@Gargron

Look into Aymara and machine translation.

Its an amazing rabbit hole.

@Gargron I've read machine translated versions of Beowulf, which I've read in various translations dozens of times. It's my favorite story of all time. And Old English, like Old Norse is a very descriptive language with hidden meaning behind it. And these machines loose those linquistic nuances. You can tell when a medieval text is machine translated.

@Gargron
Case study:

There's multicultural accessibility, and then there is the drop down language menu for Odyssey House. Included: Latin, Esparanto, and endangered european languages like frisk.

It doesn't expand the acronyms and only sometimes translates proper nouns. The lack of oceanic or Australian languages is also noted, for an Australian site. Many of the translations say ridiculous things (or so native speakers have told me).

Odysseyhouse.com.au/odyssey-multicultural-programs/

@Gargron

Also, LLMs are making machine translations worse by adding hallucinated content into the translations:

https://www.404media.co/ai-translations-are-adding-hallucinations-to-wikipedia-articles/

@Gargron funnily enough the transformer block used in most LLMs was invented for translation and TBH LLMs are much better at translation than anything before them. Not refuting your point but as machine translation goes LLMs are the best we have.

@Gargron Funnily enough, literature is an easier translation target than social media. The latter is ever evolving, the former is frozen in time.

@Gargron Stuff like this happens too.

@Gargron absolutely! Translators make a big difference to the understanding of the text. And without real, poetic fluency, the meaning would be lost.

I saw a Russian film with terrible machine translated subtitles a few years ago. I still remember a scene with the dialogue, "We produce a car!" "Release the thorns!"

I suspect it was supposed to be gritty, hardened police saying something like "Target approaching", "Deploy tire spikes".

Release the thorns does not have quite the same tone.

@Gargron This was even true in the 1600s, when the Companies of (human) Translators were translating the Bible into English (the so-called "King James" version, 1611).

Translations of human language require the ability to translate the _sense_ of some local or regional usage into something similar in the target language.

They include a footnote indicating that one passage was essentially untranslatable, because the phrase was not understood by anyone. So they used context instead.

@Gargron
I do not agree, #Garbage is a good band.

@Gargron it is weird to think the utility of machine translations is in books and film and not, say, news articles, restaurant menus, weather reports, texts to a hotel, train schedules, etc.

@Gargron

i do appreciate automatic subtitles extremely for hitting all my humor-chords. may they never evolve.

that being said:
i am lucky and able to read in several languages and read a lot of our bookclub books in original language. i can't count how many times i liked books that many of the others couldn't even finish their translated ones (assumedly) because the language was so poor. (and then we have those who listen to books and it totally depends on the person who was recorded.

@Gargron I was read the Rig Veda, machine translated from Sanskrit to English. It was great for understanding the words that were used. And after I read a human translation I understood the text, in as far as one can understand a religious text

@Gargron This.

@Gargron
For many years, YouTube has been filled with tragic machine-translated captions, even before the new AI trend of the past few years. Not only machine translated ones, but also machine generated ones converted from spoken words.

The problem with any machine translations or generated captions is that people aren't proofreading them. Viewers get garbage and the machines didn't get proper feedback.

@Gargron I worked with subtitle translations for years... I need to comment on this!

The main issue people working with machine translated subtitles is that people take models for translating things in a single modal – text – and applying to a multimodal media – video. Of course the results are horrible!

There are research on improving that, sure, I did some, even, but even we are FAAAR from getting them any good. Translating "The nurse aided the doctor take care of the patient." to many languages require guessing the gender of three people! LLMs will often default to male, female and male, due to bias.

But, the sad thing we have to admit: many works of art are so unpopular the only translations people will have are machine ones, from weird anime like Sazae-san, to Mastodon toots.

@Gargron

My main use case for machine translations is spot checking words in languages I don't know as well as I should.

They are great for that.

@Gargron As an LLM would say to a translator: "All your job are belong to us".

@Gargron Many years ago, while on holiday in Amsterdam, I bought a Dutch translation of a book by one of my favourite authors, Terry Pratchett.
In it, there was an essay, in English, by Terry, about his struggles to find a translator for the book, which was only accomplished when he realised that it wasn't just a case of taking the text and replacing it with Dutch.
No, large sections would have to be entirely re-written by the translator, to use concepts that a Dutch audience would find familiar.
And not just in Dutch, but every language.
The example he gave was one character who was experiencing the feeling of being stuck in traffic on a busy road on a Sunday afternoon, and after miles of driving, finding that the cause of the tailback was a little old lady out for her weekly drive to church in her trusty old Morris Marina, never getting above 20 MPH becuase it felt too fast.
This is something that British people are well acquanted with, but the Dutch translator had to come up with a completely different way of explaining this, because it's not something particularly prevalant over there.
It's not just about translating the words, its translating the feelings, the emotions, to something readers in another place will understand.
And LLM's fail spectacularly at that.

@Gargron

on the plus side, machine translation illuminated me as to the involvement of the presbyterian church in the famous movie "Star War the Third Gathers: Backstroke of the West"

https://en.wikipedia.org/wiki/Star_Wars:_Episode_III_%E2%80%93_Revenge_of_the_Sith#Backstroke_of_the_West

@Gargron

@Gargron I minored in linguistics in college, and a lot of exciting work was being done at the time around developing syntax models of how languages worked (and different ways humans use syntax), in part to inform machine translation models. This was more than 25 years ago. No LLMs involved.
I have not kept up with current developments in machine translation but I strongly suspect that it's built on the foundation of those decades of work actually understanding how languages function, and what maps or doesn't map. Which is completely different than expecting generative AI to create a model.

I don't get the machine translation argument. LLMs do even that poorly, certainly noticeably worse than a purpose-trained translation model, which is I believe what at least Google Translate uses.

My pet peeve about machine translation is how someone at Google and Microsoft thought that it's a good idea to "helpfully" translate developer documentation into your system language by default. As in, you have to look for a button to read it in English. Who in their right mind could possibly want that?..

@Gargron much of my work is as a legal translator (evidence, wiretaps, court filings, etc.)

The party that relies on machine translations or worse, AI translations, is the party that will lose the case. Any translator can pick holes in an AI translation big enough to cross through with a herd of elephants. Those "translations" lack nuance.

@Gargron Yeah, people who don't know anything about language or translation seem to think of translation as a perfect example of a "mechanical" process that should be automate-able.

*Maybe* for some kinds of technical writing (which still has its difficulties), but good translation of literature is probably one of the hardest things to replace humans for, right alongside writing good literature in the first place.

@Gargron

Kille Bill -> 'Kill rekening' (Kill the invoice)

@Gargron I've seen way too many machine-translated instructions, and that's something you really don't want to have machine-translated, for obvious reasons.

@Gargron totalmente de acuerdo, compañero.

@Gargron ironically, I feel translations have become worse with the rise of LLMs. For example, I recently saw a Microsoft KB article that had a table with Yes / No in its cells, translated to German as Ja / Nr.

@Gargron Denkt einfach nur an Erik Simons Übersetzungen der Witcher-Reihe (und "Etwas endet, etwas beginnt"-Kurzgeschichten). Viele Wortspiele aus dem Polnischen würden 1:1 gar nicht funktionieren. Deshalb hat Erik Simon, wie ein guter Übersetzer das tut, sich eine Menge künstlerischen Freiraums erlaubt, um den Sinn und Wortwitz des Originals zu erhalten, wo immer nötig.

@Gargron or Google's auto translated crab, Voice or text is atrocious

@Gargron
In the actual literature, pre-machine translation, it was well-established that translation amounts to authorship, with the clear example of poetry -- translating poetry into a new language is absolutely an act of authorship, not a mechanical act.

It may be somewhat less true of other kinds of works, but now we're talking about degree, not of kind.

Admittedly LLM translation is pretty handy sometimes when there's no other translation alternative, but we must be prepared for it to fail utterly at times -- like other uses of LLMs.

I have the impression that primarily anglophone people don't read as much translated literature, because so much good literature already exists in their language, so this issue may not be as familiar within that demographic. As someone who did not grow up anglophone, I can tell you there is a world of difference between a good and a bad translation even when done by humans. Machine translations are not even on the scale.