Except it isn't. LLMs and AI "translation" are built around pattern recognition, not any real understanding. They're fancy adlib machines that put what is probabilistically *likely*, not what is actually there. That's why so many AI "translations" will have random garbage thrown in, because the algorithm sees some things are commonly used together, so inserts it into the output.
Translation between languages is not some simple 1:1 conversion a machine can do. It requires human understanding, analysis of authorial intent, and picking up on subtext and context in ways a machine will never be able to do.
Just because something is understandable does not make it a good translation.
