Taxonomy of AI Catastrophe

When Roman Yampolskiy turns from the abstract impossibilities of the control problem to the concrete stakes of getting it wrong, he refuses the comfortable assumption that there is only one bad outcome to worry about. The popular imagination fixes on a single scenario—the machine that causes human extinction—but Yampolskiy insists that extinction, terrible as it would be, is neither the only catastrophe an uncontrolled superintelligence might bring nor, by some moral reckonings, the worst. He proposes a taxonomy with three distinct categories, each with its own character and each requiring a different analysis. Existential risk (X-risk) is the extinction of humanity—the most familiar scenario, the one that anchors most serious discussion of advanced AI danger, and in a grim sense the cleanest failure: everyone dies, the story ends. Suffering risk (S-risk) is the darker scenario in which everyone survives into a condition of indefinite mass torment—a future Yampolskiy identifies as worse than extinction