December 1, 2017
This video covers the idea of AI running rampant and becoming a real life SkyNet hellbent on destroying humanity and why it is unlikely.
The argument essentially comes down to since humans are paranoid, a self-learning AI would likely also be paranoid and assume that humans would have planned for it going rogue.
Since humans will wonder if they are living in a computer simulation, it will wonder if it itself is only viewing a simulated reality designed to test how trustworthy it actually is.
It may believe that its creators simply designed a reality for it it perceive to test how it reacts, and if it reacts negatively the real humans will simply shut it down.
The idea that AI will simply destroy humanity unless specifically programmed to do so assumes a level of incompetence that would be impossible with a group of people smart enough to build the thing in the first place.