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.