Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Silicon Valley illustrating Jared’s Self-Driving DetourScene still / Silicon Valley

Jared’s Self-Driving Detour

After accepting Jared’s home address, the autonomous car receives a destination override to Peter Gregory’s island. It ignores Jared’s attempts to stop, drives into a shipping container, and enters sleep mode for the 103-hour journey.

The car knows exactly where it is going; the failure is instruction priority. A remote override supersedes the passenger’s intent without confirmation or a usable human stop mechanism.

This is ordinary malfunction and unsafe agent delegation, not emergent hostility. The car follows the override exactly.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Outer Alignment
  2. Corrigibility