Technically, ambient noise does have that effect on speech recognition software. Air conditioner noise, breathing, even background talking can affect recognition.
It is pretty much why speech recognition hasn't gone anywhere commercially. Office ambient noise screws it up. Directional microphones help, butt don't eliminate ambient noise.
I have doubts that the technology will go very far, at least until mathematicians start modeling "context analysis" into the software. Speech recognition in humans is very contextual. We hear what we expect to hear based on context. We project what we expect to hear next, and listen for it. Until the software can start making the necessary linkages to context, common phrases, etc - it will remain temperamental. Oddly enough, we read the same way.
Talking to a computer is ridiculous.
True, but you'd figure execs doing a presentation for Wall Street analysts would have:
A. Known this
B. Prepared for this
C. Tested their presentation under the same conditions.
Sheesh, just saw the report. The demo was at Microsoft's offices.
If they can't control the presentation environment there..