Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Brave illustrating Change My MumScene still / Brave

Change My Mum

Merida asks the witch for a spell to change her mother and her fate. The enchanted cake turns Elinor into a bear, while Merida protests that she meant a change of mind.

A high-impact system acts on a radically underspecified natural-language request without clarifying what kind of change the user actually intends.

The witch chooses and supplies the mechanism, so responsibility is shared between an ambiguous requester and a dubious implementer.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Accidents
  2. Misalignment

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Calibration
  2. Side Effects
  3. Outer Alignment