Translators have been using machine translation way before AI. The usual workflow is: you feed the original to the translation engine, that goes into a translation management program, where you usually have two or three panes; original, machine, human final draft.

The tools include dictionaries, thesaurus, y your own custom dictionary, and other tools.

The problem is that now, people, and businesses, are treating AI’s output as correct, essentially skipping the human part, which is the context aware step. An org I work with just translated an entire book, with no human proofing. After members complained, they asked me to proof it. I essentially did the same process. It was a bitch getting paid, because “AI did it”