Your model of cognition is off. When you are thinking about an orange, you don't have an actual orange-like terabyte structure appear in your brain. You think about individual body reactions you might have to an orange, or you think about specific things you could do with an orange. Which slivers of orangeness appear is highly context dependent.

The fact that you have control over your attention... you can direct it to arbitrary things you could do or feel with an orange... that gives you the subjective impression of seeing the whole thing all at once, but that's just a lie your brain tells you. You only actually experience tiny vignettes.

You imagine your experience of "orange" is this massive huge bandwidth cognitive experience, but you only really need a handful of neurons to maintain a weak signal signifying orange.

A crude drawing of an orange in an app can trigger exactly those same neurons. As long as your mind is busy with other things, you won't even notice the difference. This is why novels work. And it's why comics work. "Understanding Comics" by Scott McCloud does a great job of showing how abstract representations of things can provide richer experiences than realistic ones.

Of course, if you direct your attention to the differences between the crude drawing and a "real" orange, you can interrogate those differences. But the fact that you can explore a rich representation of an orange in your brain doesn't mean that you do when you experience one in daily use.

I want orange to appear on a google search, what bytes need to be sent over the wire from my brain.

Wrong paradigm: imagine being able to have perfect recall of information you looked at once. Or being able to check point your active memories and recover that state of mind at will.

If I'd had to guess and theorise, and it's highly likely I'm wrong, making thoughts available semantically for an computer to parse requires more bandwidth than (once command+control is figured out) recalling of information + presenting it. The latter could be directly tied to the visual nerves only perhaps? Still all part of the same paradigm to me though.

Wrong paradigm: Imagine your brain has the entire contents of Google. The speed of all machines connected to the network. And the creativity of all other linked minds.

But it won't, unless Musk somehow figures out how to "write" information to your brain, and right now we can't even fully figure out how the information is stored, much less how to write it back - the way this will most likely work is that you will search something on google by thinking about it, and it will "appear" in your vision by stimulating your visual nerve. You still have to read it and remember it, you won't magically know Kung-Fu Matrix-style by thinking about it.

That's not substantially better then "easier" technology - i.e. something non-invasive but advanced, like contact lenses with a lightfield display and wireless power.

But if you're going to be interfacing with the brain, then there's a lot we can do probabilistically - i.e. deliberately inducing different parts of the memory centers of the brain to promote recall, targeted to the patterns of activation at previous times.

Modifying the production of neurotransmitters or being able to deliberately dampen some would also lead to some interesting possibilities.