Thinking about it, translation now seems to be a cycle, rather than just a linear process of going from Language A to Language B.
This is a good thing, as it shows the market is becoming more sophisticated and discerning. Here’s my take on it (feel free to add, disagree, go into raptures):
Spoke 1: Machine Translation (MT)
For purposes of general comprehension: Provided the text is post edited by a trained post editor/translator, it will be accurate and should contain no spelling errors. Skynet hasn’t gone rogue just yet, but it’s not bad and the Humans are still in control.
Spoke 2: Human Translation
Most common choice still. It is accurate and faithful against the original, mirrors the tone and the choice of words and if it is something like a patent or contract, can be used in a legal context.
Spoke 3: Localisation /Localization (see what I did here?)
It is culturally sensitive, highly practical (works) if it is software code. It handles elements such as local idiom, measurements and dates. But does it really speak to the target market? Ideal for scientific content.
Spoke 4: Transcreation
It literal terms it may no longer resemble the original text in a great many respects, but it is a far better product for the target audience. Ideal for highly artistic content.