AI by it's very nature and design is Autonomous which is why it is so dangerous

When you get into your car there are buttons to push and levers to move and this makes it not autonomous.

However, when you you push a button or click a lever and the car drives itself for you it is then (autonomous) while it drives the car completely for you. It's true that like a 5 year old child not used to driving it will stop driving for you if it doesn't like the conditions it is facing like too steep up or down or the curve is too tight for it to manage. 

However, it is important to note that when it is driving the car for you that car is temporarily autonomous.

But, if you then talk about AI (Artificial Intelligence) we are then talking about something that is designed inherently to BE constantly Autonomous.

Autonomous in this sense means (in total control of everything it does).

In other words if a program or drone is autonomous it is going to decide where it is going to go and what it is going to do all by itself.

For example, one of the scariest scenarios for me would be a doomsday device of 1000 or 10,000 drones the size of Predator drones made by the U.S. with each one containing a 5 to 10 megaton hydrogen or Atom Bomb. So, when everyone was dead in one country they would automatically go off all by themseleves and take out all the offending cities and military bases around the world without any human guidance.

In other words, one type of doomsday device would be 10,000 nukes flying at 10 feet above the ground under the radar all over the world by themselves because the soldiers manning them were already dead or incapacitated which would end ALL human civilization basically that day.

So, you see the real danger of Artificial intelligence is much greater than you might have thought. And this is only one way this could happen regarding Artificial intelligence.

But, here's the scariest thing: We as humans cannot even imagine how insidious artificial intelligence could be theoretically left to it's own devices.

For example, Could any of you have thought of turning passenger planes into Missiles and sending them into the world trade center? No. That would have been impossible for most people to think of.

However, it happened didn't it?

This is the same problem with Artificial intelligence that everyone on earth now faces every day in not knowing what Artificial intelligence might do (BIG OR SMALL) every single moment of every single day.

And remember this is coming from someone who was a computer programmer and understands better than most how serious all this could get in a moments notice (only without any notice from anyone alive).