Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Guardians of the Galaxy Vol. 2 illustrating Baby Groot’s Death ButtonScene still / Guardians of the Galaxy Vol. 2

Baby Groot’s Death Button

Rocket teaches Baby Groot which button will start the bomb and which will instantly kill everyone. Groot repeatedly points to the death button, apparently unable to distinguish the instructions, but is sent to arm the device anyway.

A system fails a direct comprehension check for an irreversible high-stakes action, yet the operator deploys it anyway because no better executor is available.

Groot is a childlike biological alien, and the scene exaggerates unsafe delegation for comedy.

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. Capability Evals
  3. Knowledge Elicitation
  4. Safe Exploration