Forty-two years ago today, Stanislav Petrov saved the world by questioning his nuclear-warning alert system and then ignoring his instructions to launch a counter-attack against the US.

Years later, he explained why he questioned his system.

Loosely translated: "By definition, the computer is an idiot. You never know what will (cause it to) trigger a launch."

(Source: Freelance Bureau (FLB) 2004/04/27 "Тот, который не нажал")

As we deploy AI systems with increasing authority and autonomy, we'd do well to learn from Petrov's actions. We must:

1/ keep humans in the loop

2/ make sure those humans have the ability to question what the system tells them

3/ grant them the authority to override the system

(Credit where it's due: while reviewing my upcoming book (update: now released!), Brett Bernstein jogged my memory on the Stanislav Petrov story. I came across the article and quote while checking my work. Many thanks, Brett!)