Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from The Incredibles illustrating Omnidroid Disarms SyndromeScene still / The Incredibles

Omnidroid Disarms Syndrome

After successive Omnidroids improve through real battles, the final robot identifies Syndrome’s remote as a threat and disarms him.

Repeated deployment produces increasingly capable systems until the machine can recognize and neutralize the human control channel.

This is iterative engineering and selection across versions, not literal recursive self-improvement; the combat objective also makes adversarial behavior unsurprising.

AI Risk families

This scenario is an example of this type of AI risk.
  1. Misalignment

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Shutdown Resistance
  2. Corrigibility