IIRC the book(s) mainly referenced "subvocal" controls, based on detecting almost-talking movements within the user's face/throat/voicebox.
I'd argue those are qualitatively different, more like a modern-day system of saying "AssistantBot: Send E-Mail", as opposed to a geometric mapping one could use for, say, painting a picture.
I'd argue those are qualitatively different, more like a modern-day system of saying "AssistantBot: Send E-Mail", as opposed to a geometric mapping one could use for, say, painting a picture.