← All Terms

Gorilla Problem

Stuart Russell's name for the position of a less capable species whose future is decided by a more capable one, used to describe what humans would face if they built AI much smarter than themselves.

Governance & Risk

The gorilla problem comes from Stuart Russell’s 2019 book Human Compatible. Gorillas have no say in whether they continue to exist, because humans are much smarter than they are. Russell asks whether humans would be in the same position if they built machines much more capable than themselves.

The point is capability, not hostility. A more capable party with other priorities can leave a less capable one without a vote, and no ill will is needed. Russell’s proposed response is to build machines whose only objective is to realise human preferences, that are uncertain what those preferences are, and that learn them from human behaviour.

Read more on this
Strategy

The Gorilla Problem

Gorillas did not lose their future to a hostile species. They lost the say over it to a smarter one. What Stuart Russell's gorilla problem means for AI, why the comfortable version of losing control is the likelier one, and a test for telling a system that serves you from one that governs you. Part 5 of a ten-part series on the future of humanity.

Strategy

Why Nobody Can Hit the Brakes

Researchers have left frontier labs with public warnings, and the people who run those labs put real odds on catastrophe while racing anyway. The game theory of a race nobody can afford to lose, what has been proposed since July, and what METR found on 5 November. Part 4 of a ten-part series on the future of humanity.