Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Elysium illustrating Talk to a Human?Scene still / Elysium

Talk to a Human?

An automated parole officer extends Max's parole by eight months from a police report and a personality matrix predicting a 78.3% chance of regression. It cuts off his explanation, offers a pill for his raised heart rate, and treats possible sarcasm as another offense.

A probabilistic score becomes a verdict while the nominal human-escalation option offers no meaningful recourse. It is a compact, funny example of automation replacing due process.

The system is deliberately punitive; the failure is institutional deployment and governance, not an autonomous model going rogue.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Automation Bias
  2. Calibration
  3. Algorithmic Bias
  4. Contestability
  5. Digital Repression