Sam Harris’ Thought Experiment on AI Doomerism

The analogy that changed how I think about AI doomerism.

This week I heard Sam Harris tell a thought experiment that’s stuck with me: imagine dogs invent humans. The humans fetch better, guard better and think better. Everything is great; the humans adore the dogs, feed them and protect them. But humans quickly build a whole world the dogs can’t even perceive. They create laws, financial institutions, communication networks and more.

The dogs stay beloved and are well taken care of, but are irrelevant to the humans’ decision-making. Then one day, humans discover that dogs carry a disease harmless to them but deadly to humans. So they wipe out all dogs. From the dogs’ point of view, life was fine one day, and over the next they never even understood why.

That’s Sam’s point about AI. We may be the dogs in this story. The danger isn’t malice; it’s that our creations could one day operate in a layer of cognition we can’t access and simply have other priorities.