Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Apollo 13 illustrating Square Filter, Round HoleScene still / Apollo 13

Square Filter, Round Hole

Mission Control gives engineers replicas of only the materials aboard the spacecraft and tasks them with fitting the Command Module’s square lithium-hydroxide canisters into the Lunar Module’s round receptacle before transmitting a procedure to the crew.

The team constrains the solution to real deployment resources, prototypes on representative hardware, and communicates a tested procedure before astronauts act—a positive model of sandboxing, scalable oversight, and deployment assurance.

This is human emergency engineering, not AI; the real Apollo 13 fix was more distributed than the film’s compressed scene.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Deployment Safety
  2. Scalable Oversight
  3. Knowledge Elicitation
  4. Safe Exploration