Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Mad Men illustrating It’s Toasted

It’s Toasted

Faced with evidence that cigarettes are deadly and chemically indistinguishable, Don reframes Lucky Strike around the harmless fact that its tobacco is toasted.

Persuasion can optimize what people notice and believe without improving truth, redirecting attention from the important fact to a comforting one.

This is human advertising, not automated persuasion.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Value Alignment
  2. AI Disinformation
  3. Information Pollution