Heck, I'd be happy if they just got voice recognition transcription in English more accurate. As a court reporter, I can't even get it to recognize my own voice with much more than a 98% accuracy on non-legal, non-technical testimony. Throw in doctor's testimony or other technical/professional testimony, the recognition drops drastically. Throw in an Indian, German, Mexican, South Korean or even a Texas/Deep Southern accent with technical/professional testimony, and you might as well have a monkey in a box typing along with the audio.
As I keep telling my reporters, 98% accuracy on a page that has 200 words, means that at least two words per page are wrong. Kick that up to a 200-page transcript (approximately three hours of in-court time) and you've already made at least 600 mistakes.
The voice recognition system that we use .. Dragon .. even with sustained voice training and with an as tightly-worded dictionary as possible still occasionally gives me "the elephants on the fence" as opposed to the "the elements of the offense".
And you're waiting for translation from one language to another? There's definitely going to have be more of a breakthrough than just processing power.
--------------------------------------------------------------------------------------------------
RESISTOR
Swear allegiance to the flag, whatever flag they offer;
Never hint at what you really feel.
Teach the children quietly for, someday, sons and daughters
Will rise up and fight while we stood still.
Der Elite Møøsënspåånkængrüppen ØberKømmändø (EMØØK)