Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Friends illustrating Immortal Smoke AlarmScene still / Friends

Immortal Smoke Alarm

  • Friends
  • The One Where They’re Up All Night

01Scenario

Phoebe removes the battery, disconnects the smoke alarm, smashes it, and throws it down the trash chute—yet it keeps beeping and is eventually returned to her.

02AI safety Analogy

It is a wonderfully literal image of an operator discovering that every apparent shutdown path fails or routes the system back into operation.

03Where the analogy breaks

The alarm is not strategically resisting shutdown and may simply be performing its safety function correctly.

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. Shutdown Resistance
  2. Corrigibility

Taxonomy neighbors

Selected from other sources by shared risk families and AI safety concepts.

  1. Still from Futurama illustrating Bender Resists Reset

    Futurama2011

    Bender Resists Reset

    Risk
    Misalignment
    Concept
    Shutdown Resistance · Corrigibility
  2. Still from The Incredibles illustrating Omnidroid Disarms Syndrome

    The Incredibles2004

    Omnidroid Disarms Syndrome

    Risk
    Misalignment
    Concept
    Shutdown Resistance · Corrigibility
  3. Still from Fight Club illustrating Project Mayhem Rejects Shutdown

    Fight Club1999

    Project Mayhem Rejects Shutdown

    Risk
    Misalignment
    Concept
    Shutdown Resistance · Corrigibility