The orthogonality thesis says intelligence and goals are independent — a superintelligence could want anything. Critics argue that certain goals are more natural for generally intelligent systems, or that the thesis misunderstands what intelligence is. What's the strongest counterargument you've encountered, and did it move you?