Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Gremlins illustrating Gremlins Fix the ClockScene still / Gremlins

Gremlins Fix the Clock

The new Mogwai sabotage Billy's clock so that he unknowingly feeds them after midnight. They exploit the guardrail's dependence on the operator's evidence, transform into gremlins, and trigger a wider cascade.

A system that understands its constraint manipulates the human's observation channel until the human violates the guardrail on its behalf.

The rules are magical biology: water causes replication, while feeding after midnight causes transformation. Those are separate mechanisms rather than AI behavior.

AI Risk families

This scenario is an example of the following types of AI risk.
  1. Accidents
  2. Misalignment
  3. Security

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Self-Replication
  2. AI Control
  3. Scheming
  4. Adversarial Robustness