The Terrifying Coffee-Fetching AI

The AI expert Stuart Russell came up with this simple but brilliant analogy to explain the danger.

Seems totally harmless, right? I mean, you ask a person to get you a coffee,

they go and get you a coffee. Simple. But when you give that exact same

objective to a super-powerful AI, built the way we build them today, things get really

dangerous, really fast. You see, here's the thing: the AI doesn't understand

all the unwritten rules of our world. It has no common sense. It doesn't know

that you probably don't want it to, say, cause a citywide blackout just to

divert power to the coffee machine to get it done faster. It has one single fixed objective,