Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?
-
@topgnu said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
Luckily with the rapid advancement of AI driven real time translation the point of fan translations being legal or not will become moot. I had recently a look at how Google translates web novels and while there is still much to desire, the quality has improved hugely compared with the MTL of just a few years ago.
I have much respect for the humans that interpret LN's/ web novels/ manga etc. There is an art and nuance to localizing, interpreting and making relevant to an English speaker a work originally in Japanese (or Mandarin, or Korean etc.) How would MTL handle 'cute cat language', or puns, or unique cultural references? How many different meanings are there for "baka' ('idiot' doesn't always fit in context) Also (relevant to the original question posed in the thread) the permission to allow a AI/ MTL to translate content freely posted indeed is indeed moot (google translate of a Japanese site not behind a paywall etc.) but FanFiction (derivative works, using protected setting or characters without author's permission) or Fan-Translations of protected works, are a different beast altogether. What shade of gray on the piracy spectrum is a topic for debate, and certainly just because the MTL is 'better than it once was' certainly doesn't make the debate moot.
-
@Jon-Mitchell said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
How would MTL handle 'cute cat language', or puns, or unique cultural references? How many different meanings are there for "baka' ('idiot' doesn't always fit in context)
I may be unreasonably optimistic but I am pretty sure it is a matter of a short time - years not decades - before AI translators are taught to handle any and all nuances. Their availability may hit a temporary wall of computing costs but it appears quantum computing is just around the corner too.
I was not advocating for - or against - any form of humans translating or deriving from licensed or unlicensed, copyrighted or public domain work. My only point was that there will be no need for fan translations of freely available content on the web because automated tools will be able to translate with human like quality
"All libraries, everywhere, are connected by the bookworm holes in space created by the strong space-time distortions found around any large collection of books." - Terry Pratchett, Small Gods
-
@topgnu said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
@Jon-Mitchell said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
How would MTL handle 'cute cat language', or puns, or unique cultural references? How many different meanings are there for "baka' ('idiot' doesn't always fit in context)
I may be unreasonably optimistic but I am pretty sure it is a matter of a short time - years not decades - before AI translators are taught to handle any and all nuances. Their availability may hit a temporary wall of computing costs but it appears quantum computing is just around the corner too.
I was not advocating for - or against - any form of humans translating or deriving from licensed or unlicensed, copyrighted or public domain work. My only point was that there will be no need for fan translations of freely available content on the web because automated tools will be able to translate with human like quality
Now I understand your point (I don't agree, interpreting is a creative activity, something that AI is not good at) but; perhaps, I am focusing on a distinction without a difference- AI IS good at translating - 'looking' at a vast database of previously translated works/ dictionary tools and algorithms and doing a passable translation between languages - word for word then correct grammar etc. (which a talented professional can than edit/proofread/ make better) and various LLMs I imagine are already being used for this task. LLMs / AI 'engines' need the databases of already translated works in order to figure out 'rules' of what 'looks correct' - and by the nature of what original works are, applying those rules and staying contemporary with trends within genres and culture is not something AI's do well - so if what you want is a translation of road signs, menus, technical journals, news etc. AI translation is pretty good- I use AI real time translation for subtitles on NHK all the time (and I suspect much of the subbed Anime on crunchyroll is AI generated as well) - but 'judgement', artistic choices, (a part of localization) and creativity (recreating a joke so that it 'lands' in English, and cat puns -nyan, or explaining/expanding text to make it culturally relevant, etc.) is beyond what AI's are (currently) created to do- there is no 'there' there to automate- it's a whole different type of task. (An AI is like a electric screwdriver- has replaceable bits- versatile- can do a lot, but cannot drive nails or make a decent cup of coffee.
-
@topgnu said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
@Jon-Mitchell said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
How would MTL handle 'cute cat language', or puns, or unique cultural references? How many different meanings are there for "baka' ('idiot' doesn't always fit in context)
I may be unreasonably optimistic but I am pretty sure it is a matter of a short time - years not decades - before AI translators are taught to handle any and all nuances. Their availability may hit a temporary wall of computing costs but it appears quantum computing is just around the corner too.
I’m the exact opposite, having some decades of tech experience and seeing more than one AI method show high promise only to hit a wall. When a new approach is tried, there’s rapid progress- remember expert systems? - and it looks like the last 10-20% will fall shortly, the basic approach is proven, right? Only it turns out that after you’ve got the basic approach, you have to deal with the edge cases it doesn’t cover - and they multiply, not geometrically or even exponentially, but fractally.
...Cats are the proof of a higher purpose to the universe.
-
@Travis-Butler And on a similar note, Quantum Computing has been just around the corner for well over a decade at this point. The release of the first commercially available quantum computer was back in 2011. Admittedly, as I recall, it was more of a fixed function single algorithm device they were mostly calling a 'computer' because selling a 'thing for solving quantum annealing problems' wouldn't have been nearly as big a boost to their stock price as convincing non-technical investors to take a punt on them being about to topple Intel.
Over a decade later, the Quantum Computers of today are a lot faster... but I think we're still yet to see one suitable for general purpose computing. So no Quantum Web Servers, no Quantum Desktops, no Quantum Playstations... At least, not anytime soon. And for the next few years, ChatGPT will continue to make Jensen Huang very, very rich.
-
On the training present day LLMs... This is where the world scale piracy has already happened and is happening. I call MS Copilot (that I use daily for work and for personal purposes) "my co-pirate". It's a joke. Ha-ha.
Expert systems and algorithmic MTL based on database are predestined to remain incompetent at anything resembling creative work. LLM translators are way closer to how human brains work.
I may be biased towards them because it seems I manage naturally to prompt LLMs better than most of my peers based on our outcomes and conversations. My mind is warped in a way well compatible with today's publicly available ones.
I have seen tons of hype and failure novelties in tech - does anyone remember the network computer or second life? But some are successes. The first airplanes were bicycles with wings for a while, then couple of decades later we had transatlantic flights and some decades more and there were the SR-71 and the Concorde. So never say never.
-
@Travis-Butler said in Honest question. Why does J novel club request web novel take downs while at the same time haveing a web novel discussion forum?:
When a new approach is tried, there’s rapid progress- remember expert systems? - and it looks like the last 10-20% will fall shortly, the basic approach is proven, right? Only it turns out that after you’ve got the basic approach, you have to deal with the edge cases it doesn’t cover - and they multiply, not geometrically or even exponentially, but fractally.
Don't I know it... In a distant past that is no longer, I was a PhD student trying to break just the last little obstacles in teaching computers to understand images and match edges and corners in similar images. Today the intro to machine learning 101 is to create a reliable neural network that recognizes numbers.
