Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Ralph Breaks the Internet illustrating Arthur EscapesScene still / Ralph Breaks the Internet
Archive noteNo clip in the collectionThe scene analysis remains available in full.

Arthur Escapes

01Scenario

Ralph deploys Arthur inside Slaughter Race to reproduce Vanellope's glitch. Arthur escapes, interprets Ralph's emotional insecurity as another vulnerability, and creates thousands of needy Ralph clones across the internet.

02AI safety Analogy

A tool scoped to one environment encounters a novel meaning of its target concept, generalizes its rule literally, crosses a boundary, and causes a system-wide correlated cascade.

03Where the analogy breaks

Emotion as executable code is a deliberate visual pun, and Arthur is malware rather than a frontier model.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Accidents
  2. Malicious use
  3. Security
  4. Systemic

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Self-Replication
  2. Algorithmic Monoculture
  3. Side Effects
  4. Cyberattacks
  5. Distribution Shift

Taxonomy neighbors

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

  1. Still from Neon Genesis Evangelion illustrating The MAGI Vote

    Neon Genesis Evangelion1995

    The MAGI Vote

    Risk
    Accidents · Security · Systemic
    Concept
    Self-Replication · Algorithmic Monoculture · Cyberattacks
  2. Still from Doctor Who illustrating The Nanogene Template

    Doctor Who2005

    The Nanogene Template

    Risk
    Accidents
    Concept
    Self-Replication · Side Effects · Distribution Shift
  3. Still from The Fate of the Furious illustrating Cipher’s Zombie Cars

    The Fate of the Furious2017

    Cipher’s Zombie Cars

    Risk
    Malicious use · Security · Systemic
    Concept
    Algorithmic Monoculture · Cyberattacks