I'm pretty sure crowdsourcing subtitles in different languages is the primary use of the feature precisely because machine generated translation of machine generated captions produces gibberish.