Cyrus 🌎

Since we're doing AI risk again, people are questioning how AI could actually physically kill some or all humans. This is a bit like people in February 2020 questioning how a relatively small virus outbreak could bring the whole world to a halt and kill tens of millions of people. People are, in general, Trend-Blind. They can see where things are now, but they can't extrapolate. At all. AI today can solve math problems that are roughly at the level of the best human mathematicians in the world. 5 years ago, it could babble like a toddler. Ok, that's the trend, but where is it heading. In 5 years time, it's plausible that AI will be able to mount serious long-term espionage and cyber operations autonomously, and also run its own covert installations in various places around the world. It may be able to influence politics, media and global events. At the same time, the world of 2031 may have a lot of robots - cars, drones, robotic security, military drones. Could AIs in such a world go a head and "delete" all the humans? Yes, I think so. And the way that the final blow falls might not be that interesting - perhaps everyone's phone blows up because the factories that make phones were somehow coopted into putting explosives into the battery. Perhaps the drones that the police and military use will be hacked and go blow people up like they do to soldiers in Ukraine. Starvation could be used too. Take down the services that keep society running and selectively destroy any coordinated attempts to fix them. People don't understand how their own civilization works The key difficulty for AI takeover isn't how to kill people (that is easy). It is how to take control of the systems that run society - politics, media, academia, finance, law enforcement, military - or at least prevent those systems from responding properly to attacks. And those societal control systems that we rely on are not at all secure. They are all hackable. In fact they have all been hacked in various ways by various merely human entities. Some may even claim that humans have made the control system insecure on purpose so that elites can more easily extract value from society. Either way, that societal control system is what AI will attack if it wanted to end humanity.