"generative models are fine"say you have no friends without saying you have no friendsif you had friends who are visual artists, musicians, writers, translators, or any other affected profession, your views would be different
-
"generative models are fine"
say you have no friends without saying you have no friends
if you had friends who are visual artists, musicians, writers, translators, or any other affected profession, your views would be different -
"generative models are fine"
say you have no friends without saying you have no friends
if you had friends who are visual artists, musicians, writers, translators, or any other affected profession, your views would be differentSo, in your opinion were machine translators always a bad thing? Genuine question.
-
So, in your opinion were machine translators always a bad thing? Genuine question.
@cloudskater no, in case of genai i feel the problem with translation specifically is over-reliance. they still strip the humanity off everything
many people also do not understand the intricacies of translation and how nuanced it all is
good translations involve way more than just replacing words -
@cloudskater no, in case of genai i feel the problem with translation specifically is over-reliance. they still strip the humanity off everything
many people also do not understand the intricacies of translation and how nuanced it all is
good translations involve way more than just replacing words@lumi @cloudskater I mean LLMs are also just worse for translation in many ways than the machine translations we already had prior. they have the same issues of not understanding context and nuance, but not just are they much larger models with more power use, they also are prone to just making stuff up too. which is not something you want in a translation. at the same time the output might look more polished, giving the impression of a better translation while sacrificing accuracy. I think that's actually very dangerous.
-
@lumi @cloudskater I mean LLMs are also just worse for translation in many ways than the machine translations we already had prior. they have the same issues of not understanding context and nuance, but not just are they much larger models with more power use, they also are prone to just making stuff up too. which is not something you want in a translation. at the same time the output might look more polished, giving the impression of a better translation while sacrificing accuracy. I think that's actually very dangerous.
old machine translation didn't have context and sometimes used wrong words or didn't get the grammar right
llm based translation still doesn't have context but acts like it does anyways and is prone to just making up completely new shit that wasn't in the source
old machine translation was bad in predictable and noticable ways
llm translation can sneak in hard to spot mistakes and is extremely unpredictable
-
old machine translation didn't have context and sometimes used wrong words or didn't get the grammar right
llm based translation still doesn't have context but acts like it does anyways and is prone to just making up completely new shit that wasn't in the source
old machine translation was bad in predictable and noticable ways
llm translation can sneak in hard to spot mistakes and is extremely unpredictable
@elexia@twoot.site @lumi@snug.moe @sammy@cherrykitten.gay
Are there any translation programs left that don't use LLMs that you know of? Preferably open source.
-
@elexia@twoot.site @lumi@snug.moe @sammy@cherrykitten.gay
Are there any translation programs left that don't use LLMs that you know of? Preferably open source.
-
R relay@relay.an.exchange shared this topic
