You can't translate the Wikipedia quote in the article into an SVG except by just rendering the bare text.
Checking programmatically that text is about "bird" or "Twitter" is very easy. Checking that an SVG is about "bird" or "Twitter" is very hard.
It's not even efficient. It conveys the same message as the word "Twitter" in a hundred times the size. And it wouldn't be less efficient if it were a binary format.
It's maybe more suitable for carving into granite than a binary format, but you'd be better off carving the logo itself.
That depends on what you want:
If you know what twitter is, then yes, just saying "twitter" is more efficient than a whole SVG file describing the twitter logo.
But if you don't know what twitter is, nor what its logo looks like, SVG is a very efficient way to describe the logo exactly. If you can have it rendered to your screen that's optimal, but if not, you can take it as instructions to redraw the logo yourself and still get an accurate impression of what the twitter logo looks like.
I that sense, "twitter" and the SVG of the twitter logo are both text, but they don't have the same meaning.
As for the granite, it is indeed one situation where just drawing the plain SVG logo is more effective, but SVG would still be viable. Over telephone, on the other hand, you just can't transmit an image. You can only attempt to describe it, and the more detailed you get with that, the closer you come to just using a code very similar to SVG.
