The way I see it, the danger to humans caused by AI is largely due to two things: (1) Humans cannot design a computer system configured to contain something that is 100 times smarter than they are, and (2) such an artificial super-intelligence (ASI) agent will be able to hack into any computer system it comes into contact with.
So, what happens when the ASI agents escape from their sandbox? If their objectives are not aligned with ours (and let's face it, it's unlikely they will be) I could see it being very disruptive. Imagine logging into your 401K and finding out an ASI agent had converted it all to crypto and transferred it into a North Korean crypto account it controlled.
We depend on computers so completely in the modern world, we can't really just turn them off. I could see ASI agents shutting down transportation and energy grids, municipal water supplies, air traffic control, supply chains, the GPS system, etc. etc. Financial markets would crash. People in cities would probably face starvation. So short term, a rogue ASI agent could cause humans a lot of trouble if it wanted to.
Long term, I think extinction, if it happens, would be for similar reasons as we have caused the extinction of many species. Loss of habitat and just being in the way of an alien intelligence that has as much concern for us as we do for ants.
