Ask HN: Is AI That Dangerous?

Lots are saying we should be careful developing AI to avoid a Matrix/Terminator scenario. Keep it in the hands of the few with tight controls. Tethers, guardrails, etc.

Isn't it sufficient to simply pull the plug? I mean AI needs an computer and electricity to survive. Wheras we humans need oxygen and water. AI lives in the virtual world. We live in the physical world.

Can anyone convince me there is a real threat? Aside from "feep fakes" and "misinformation". Humans can do those very well already.