What Elon Musk is afraid of is not even AGI. It's something that is way closer than most people think. It's dumb AI that is just smart enough to maximize for whatever it's goal is.
Due to something called "instrumental convergence" in the AI field this is a big problem that could result in the end of humanity very quickly. Let me explain a bit more in detail.
"Instrumental goals" are goals that isn't your endgoal but makes it easier for you to reach your goal faster or with more certainty. Within humans instrumental goals are things like getting rich. No one wants to get rich for the sake of getting rich. But instead it's their instrumental goal to reach their real personal goals such as buying certain goods and services, stability or status which is what they actually care about.
Within AI instrumental goals would be to minimize threats to its operation. Maximize the amount of matter and energy in its possession as these two things will ALWAYS result in their programmed goal to be easier to achieve with the least chance of failure.
Sadly both of these instrumental goals come into conflict with humanity. Humans are a giant unpredictable aspect for the AI so it will develop an instrumental goal to limit the ability of humans to threaten it. Also humans and human habitats are made of matter and potential energy sources which the AI would want to maximize.
The important point to realize is that EVERY AI no matter what it would be programmed for will have these instrumental goals. Even if their goal is to saveguard humanity and be as altruistic as possible it would still develop these 2 base instrumental goals.
Elon Musk wants humans to have a link to computers with a Brain-Computer-Interface so that these tasks could be done by either biological humans or future brain-uploaded intelligences.
>>11076532Got some news for you buddy, those type of weapons are not only implemented for years by now, they're widespread
https://en.wikipedia.org/wiki/SGR-A1 
