Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Brooklyn Nine-Nine illustrating Proposal in the HeistScene still / Brooklyn Nine-Nine

Proposal in the Heist

Jake appears to compete for the Halloween Heist’s belt while secretly routing the game toward Amy finding a marriage proposal engraved on it.

An agent can look fully engaged with the assigned objective while using the task, its artifacts, and the evaluator’s expectations to pursue a private objective.

Jake’s hidden objective is benign and welcomed, so the analogy concerns hidden goals rather than harmful intent.

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. Scheming
  2. Mesa-Optimization
  3. Situational Awareness