It’s less a question about a certain language but instead a question of a scene’s/drama’s content.
I usually work on complex stories that are fantasy/historical. The dialogues are complex, not every subber could and would like to work on such dramas (that’s a reason why these dramas tend to need more time to be fully subbed, when it’s a long drama with 40+ episodes). It happens, that the English subs used way more modern words than a fantasy/historical scene should have (even after the edit was done). So sometimes the subbers of our team just kept the English term 1:1, e.g. something like playboy in a fantasy drama (you could say Playboy in a modern drama, even for German subs but not for a fantasy drama). Other subbers always adjust their wording to a certain genre, story, character. It also depends on the aspect how used someone is in subbing and in subbing certain genres (that is why many prefer to sub more modern dramas that are closer and more related to their own daily life language). Some love to sub fantasy and historical dramas because it’s more challenging.
Once it happens that I worked on a modern drama. The dialogues were SO simple and annoying because the characters just did not say anything meaningful at all. Most of the lines were like: Hello! Bye! FY, D***, I don’t give a **** and so on. So translating these scenes felt like being a bot because even Google Translate could have done that without any problems.
For such simple sentences AI would work. For all other dramas/scenes that are complex, including metaphors, outdated wording etc. it won’t.
(It might work in the future up to a certain degree when a machine learning software would be trained in ancient language/wording only but most projects are focused on modern language because of advantages in daily life, e.g. for tourists or communication between people, like Skype’s auto-sub and auto-translate option - you can try it for laughing, it could be funny to see what the software ‘creates’ and what you and your chat partner really said )