Ask HN: How is AI actively a threat to humanity if it's only online?

That's always been something that's left me confused here. Yes, an AI can be dangerous by hacking websites and company systems or flooding social media with scams or fake news. Yes, AI generated code can have security issues that a developer doesn't catch because they didn't review it properly.

But isn't there a pretty big jump from that to "can destroy humanity"? Like say, a jump involving a society not run by the internet?

Because if an AI takes over the systems at a factory, it can't really do much beyond what the factory could already do. The production lines use single purpose robots and machines to build a particular product, and can't be programmatically altered to build anything else.

So, what exactly could a super intelligence even do with access? Only things I can picture are using it to try and hack another online system or speeding up/slowing down production rates.

Similarly, I hear a lot about these systems taking over weapons or releasing bio-engineered viruses or what not, and my first thought is "how?"

These systems are not accessible via the internet. There's no API endpoint that can launch nuclear weapons or set up a drone strike via a POST request.

To me, it feels like all these apocalyptic scenarios only make sense if a percentage of humanity gets manipulated by AI into doing its bidding, or if humanity suddenly invents easily hackable intelligent robots with an internet connection.