For me, it's a question of appropriateness, not accuracy. 19 times out of 20 I want to input text simply because I don't want to speak out commands around other people. Speaking those commands is not only socially awkward (which, maybe can change) but confusing to those around you (which really can't change). Even when the social awkwardness is slightly masked by doing it through hands-free/headphone-mic/phone-to-head, I still often don't want to broadcast to everyone around me what I want to text to a friend.
In terms of technology, what's not ready yet on voice control is a way to magically read into the user's gait. That is, I often find myself pausing to think when I type, but when I do that as I'm speaking a sentence, it's often interpreted as me being done with the voice input. Similarly, if I'm speaking out a chain of directions "text Ronald McDonald that I'll be at McDonald's in 10 minutes" there's often confusion between which portion of the voice input belongs to which command. This only gets worse as the commands get more complicated.
Bottom line: I don't necessarily agree with the author that voice input "really sucks" due to input accuracy, but I do think that any experience that relies solely on voice input for interactions (especially when it's designed to be used primarily in public spaces) will really suck.
In terms of technology, what's not ready yet on voice control is a way to magically read into the user's gait. That is, I often find myself pausing to think when I type, but when I do that as I'm speaking a sentence, it's often interpreted as me being done with the voice input. Similarly, if I'm speaking out a chain of directions "text Ronald McDonald that I'll be at McDonald's in 10 minutes" there's often confusion between which portion of the voice input belongs to which command. This only gets worse as the commands get more complicated.
Bottom line: I don't necessarily agree with the author that voice input "really sucks" due to input accuracy, but I do think that any experience that relies solely on voice input for interactions (especially when it's designed to be used primarily in public spaces) will really suck.