People who are deaf, are hard of hearing, or who are having trouble understanding audio information for any reason can read the text presentation or (in the future) have it translated and presented as sign language by assistive technology.
No such vapourware technologies of text-to-sign translation will be available and reliable during the lifetime of WCAG 2. (If you think otherwise, here's something for you: Can you make it work in Quebec Sign Language?)
Captions may be generated using real-time text translation service (stenographic or, in the future, speech-to-text with corrections).
Who's gonna do those corrections (in real time, no less)? And at what point in the "future" will this be developed? The existing technique of speaker-dependent speech recognition, known as voicewriting, could be included here if defined properly.
We have updated these notes as follows:
"People who are deaf, are hard of hearing, or who are having trouble understanding audio information for any reason can read the text presentation. Research is ongoing regarding automatic translation of text into sign language."
"Captions may be generated using a real-time text translation service." (removed parenthetical)