I don't think they would use Speech to Text for preparing the transcipts. Speech recognition is still not that mature a technology to understand the huge domain of grammar.
Bill Gates in the video interview said that in 2-5 years the magic threshold would be reached.