Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from The Wire illustrating Xerox Lie Detector

Xerox Lie Detector

01Scenario

Detectives tape a suspect’s hands to a photocopier preloaded with “TRUE” and “FALSE” pages and convince him the office machine can read lies; he confesses.

02AI safety Analogy

Users may grant epistemic authority to a system they do not understand, making the appearance of automation as influential as real capability.

03Where the analogy breaks

The machine is deliberately fake and the police exploit the suspect’s misunderstanding.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Automation Bias
  2. Synthetic Fraud
  3. Interpretability

Source continuation

More from The Wire

View all 3 scenarios
  1. Still from The Wire illustrating Teaching to the Test

    The Wire2006

    Teaching to the Test

  2. Still from The Wire illustrating Hamsterdam

    The Wire2004

    Hamsterdam

Taxonomy neighbors

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

  1. Still from South Park illustrating ChatGPT Writes Stan’s Texts

    South Park2023

    ChatGPT Writes Stan’s Texts

    Risk
    Systemic
    Concept
    Automation Bias · Synthetic Fraud
  2. Still from The Wizard of Oz illustrating Toto Pulls Back the Curtain

    The Wizard of Oz1939

    Toto Pulls Back the Curtain

    Risk
    Malicious use
    Concept
    Synthetic Fraud · Interpretability
  3. Still from Atlanta illustrating Cardboard Drake

    Atlanta2018

    Cardboard Drake

    Risk
    Malicious use · Systemic
    Concept
    Synthetic Fraud