It only takes a vague input of a problem, and too much trust in the AI to maximize the goal and minimize the error. Issues start when the solutions [suggestive chatgpt, battle strategies, load balancing systems, administrative criteria] are followed blindly and unintended consequences appear afterwards. Although transparency also may create some doubt in political decision-making it’s necessary for checks & balances and even accountability if things go too far. When AI gets globally controlled by one, or maybe two large monolithic shareholder companies who decides what gets filtered, what data gets used, what is right and wrong then thus is simply too much power in the wrong hands. You get a corporatocracy dystopia on a global scale.

Lesswrong has been posting about AI doom for quite some time, they have a lot of good reads. https://www.lesswrong.com/

Billionaires don’t need AI to destroy humanity and our civilization, but they sure will it use that as well.

Show me a computer that can power and maintain itself and I’ll start worrying.

The risk AI poses is the same as fossil fuel or nuclear weapons, human decisions are the problem.

But the cool new thing is because humans struggle to understand the difference between a calculation and a decision, we can blame AI for any harmful decisions we make based on its calculations.