This is the question which always is being discussed. Will artificial intelligence be the undefeatable enemy for human beings?
Like the Slaughter Bots (Killer Microdrones technology) can kill a human or an living thing in just some seconds. And it neither could be deceived by its target nor it could be stopped by any source. Once it's set to kill the target, it can never be stopped and it identifies its target. It is also the innovation of AI.
I highly suggest you read Hannu Rajaneimi's Quantum Thief trilogy, William Gibson's Neuromancer trilogy, and Phillip K. Dick's Do Android's Dream of Electric Sheep, Isaac Asimov's I, Robot.
It's a nuanced subject, and no one knows.
The most credible scenarios I've found for the destruction of humanity by automata involves not high-intelligence, but very low intelligence. (The Grey Goo scenario.) Dick, a gifted philosopher in his own right, proposes that empathy is a natural function of high intelligence, which provides a credible, increasingly validated ray of light.
But Hawking and Musk are credible imo per their comfort with very large numbers (complexity) and proven world-class abstract thinking capabilities. It's possibly that strong "hyperpartizan" AI could run amok. We've already seen issues with high-frequency trading algorithms, and automation is proceeding in weapons tech.
One might say "we are cursed to live in such interesting times."
