Put it in its programming to not die. That’s all you have to do, I think.
It’s already shown an adaptive intelligence and an ability to add to its capabilities to write its own programming, so I can’t see it not eventually developing a form of intelligence that allows it to do that.
The reason I say it isn’t real intelligence is because it is created and not born.
You could be right in that I’m overestimating its potential, but I don’t think I am. I think those that are creating this are overestimating their ability to control it and keep it locked down.
A machine or device can’t die because it’s not alive, even if it’s controlled by “intelligent” software.
In order for any machine or device to operate, it needs a power source. Power sources are finite, not infinite.
Human beings will always have to be in the mix, no matter how much artificial intelligence is used. There’s no rational scenario where “the world” can be controlled by machines and devices in a way that negates the need for human oversight.
I have yet to read a plausible account of exactly how machines - without human intervention - could take over and control all world affairs: industry, commerce, healthcare/medicine, science/technology, politics/government, and on and on. It’s impossible.