Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Inside Out illustrating Joy Excludes SadnessScene still / Inside Out

Joy Excludes Sadness

Joy treats happiness as Riley’s overriding objective, confines Sadness to a chalk circle, and tries to keep her away from the controls and core memories.

Optimizing one appealing value suppresses an essential but initially misunderstood signal; Riley functions worse until the system incorporates sadness.

The emotions are personified mental functions, and Joy eventually recognizes and corrects her mistake.

AI Risk families

This scenario is an example of this type of AI risk.
  1. Misalignment

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Value Alignment
  2. Goodhart’s Law
  3. Side Effects
  4. Outer Alignment