I agree with your skepticism about voice becoming generally dominant, but it’s already very useful. It may also become the dominant form of usage for some systems.
It's also hard to imagine sound as a dominant interface because all we have are mediocre examples. We have to work within clunky command boundaries, rephrase commands, be in a quiet environment, not have an accent, etc.
I'm glad we're making progress, but I'll be a skeptic until I can give voice requests as naturally as I'd give them to a human. IMO there's no limit from there.
I'm glad we're making progress, but I'll be a skeptic until I can give voice requests as naturally as I'd give them to a human. IMO there's no limit from there.