It's no secret that Elon Musk (and others) are afraid of the huge, possible pitfalls of AI. And yes, usually what happens is something we never considered, sort of along the lines of "No one expects the Spanish Inquisition.", not that I think the Inquisition was at all like a Monty Python skit, but there are reasons to fear both "minded" or "mindless" machines. The irony is that they can be both simultaneously, thanks to AI.
So, Musk has come (it appears) to have come to Dr. Strangelove's conclusion: "How I learned to stop fearing and love the bomb". Musk has proposed human machine interface directly to the brain. He is proposing that interface for good, namely for stroke, cancer and congenital problems of tissue damage and replacement of lost function. Timeline? 8-10 years. For the interface itself, four years. He also believes this will grant humans the ability to communicate more quickly. Great. Facebook, and Kim Kardashian with direct access to the brain. End of privacy and taste forever.
Also, why am I tempted to say four yarons? Right. The Cylons. The infinite ability of humanity to misuse good things.
How else can one grant the machines a conscience, the knowledge of right from wrong? Problem is, who's to say that granting this won't also grant the freedom not to choose good over bad? They will have access to both good and evil, after all.
I'd say if one can limit it to therapeutic uses, it could be a good thing. However, since when can science and technology be limited?
Also, where is all this bandwidth going to come from?
I had a graphic of the plan, but photobucket is not cooperating, so you were spared, it seems.
Sources:
http://time.com/3614349/artificial-intelligence-singularity-stephen-hawking-elon-musk/
https://www.engadget.com/2017/04/21/elon-musk-expects-to-have-a-brain-machine-interface-in-four-year/