I don't make the apocalyptic predictions about AI that everyone else seems to. But there is a simple fact that anyone in IT has run into. Google (as an example) has a massive amount of money and spends it extensively to try and prevent fraud and issues with their various services.

In the process, there is story after story of someone losing their entire Google history, being blocked from Google Ads, Mail, etc. all because some automated system kicked in and decided they were a problem. Since Google doesn't have any reasonable appeals process this usually turns into a kafkaesque hell for people who experience it.

Now imagine you put that exact same computer system in control of sniper rifles and drones that can kill someone.

Yeah. It's not an ethical dilemma, it's a hellish reality.

Let's wait until Google (or someone with essentially an unlimited supply of money) can figure out how to make it so that they can run their own services without randomly killing people digitally before we start to kill people in real life with those same tools.