What's the use of voice recognition software if I still need to use my fingers to correct wrongly chosen words? I bet it would be faster if I typed the words myself!

This capability is already included in all of the speech recognition software that I know of, and has been around for decades.

This only makes sense in the 1990-to-2010 paradigm that has businessmen for some reason doing all their own typing.

Far simpler and quicker than a speech-recognition interface would be a person, trained to either touch-type or write shorthand while you spoke. We could call them "secretaries".

If you can speak neatly, this software saves you a lot of time as corrections are relatively few.

Secretaries are not slaves: this software is just for them, obviously not for top managers, who always have who makes grunt work for them!

This is actually very similar to the way that Japanese language text is entered on a standard keyboard. Because homophones are so common, as the user enters a phonetic representation of a sentence, the input method software guesses the appropriate Kanji (Chinese character) representation. If the user rejects the proffered representation, the software displays its less-favored guesses.

Speech recognition is the technology of the future... in the future somebody will make money with this stuff... you can go back 10, 20 years and find a similar "we almost fix this thing" article.