Quote Originally Posted by GrizzlyTank View Post
Although Google translate, deepl and probably some other translators using the newer generation neural network models have gotten surprisingly good.
Alright, I'm getting technical here!!!

They'll still work off of corpora. Which is basically this large collection of texts in both languages where lines and sentences are connected. It's how those neural network models grew and will continue to grow. The more they're fed by Google itself and us users, the better they will get. It's why I say "careful with what you put into Google", because they will store it into their database. From the silly stuff to the really important "I kinda shouldn't be giving this to anyone" stuff.

And I won't deny that they've gotten really good. In fact, even when I was still in my course, my teachers were already saying "eventually Translators will be doing revision work more than actual translation work". Which is still valid; cultures and the emotional aspect of humans are just so complex that you can't expect a machine to understand every single bit of nuance possible. You'll only have Google Translate be 99% reliable for technical texts like science papers or instruction manuals. And even then, you'll still need a human to review it. So always hire someone to at least make sure you got proper quality. You won't fully replace translators, but you will change how they do their job.

But otherwise, you are correct, and I think there's far less stigma around using Google Translate and machine translation in general thanks to how advanced they've become.