Maybe in some future sci-fi world like Star Trek with its universal translators, but no one seriously believes that computers will be able to deliver rapid and context-specific translation in any future that we can reasonably foresee. For certain applications, like translating newspaper articles, we can reasonably expect pretty darn good translations within a decade, but for many uses, particularly conversation, we can’t expect computers to be anywhere near as good as humans.
Polysemy will simply defeat the computers. For example, I was just translating a medieval saint’s life from Latin and encountered the word miracula. This can be miracle, or it can be wonder, amazement, wondrous [thing], freak, freakish [person], or hideously ugly woman (plus probably a few other meanings I can’t think of). Does praedico mean to proclaim, to publish, to praise, to preach, to warn, to predict, to recommend, or to say beforehand? Which sense best fits a given context is not something a computer is going to be able to determine well. It can probably get pretty close, but for a really fine translation, you’ll need a human.
Then there is translating handwritten documents and manuscripts.
And then there’s translating poetry and other literary works. That is an art and requires aesthetic judgment on the part of the translator. Not to mention that no matter how good the translator, it will not be same as the original. For example, Seamus Heaney is among the greatest of living poets, and he is an excellent scholar of Old English. But while his translation of Beowulf is a fine one, it just doesn’t compare to reading it in the original. The poem is so much more subtle, beautiful, and powerful in the original Old English.