Cultural Alignment

Exploring real AI risks through the lens of pop culture

Still from Anatomy of a Fall illustrating Three Blood Drops, Two WorldsScene still / Anatomy of a Fall
Archive noteNo clip in the collectionThe scene analysis remains available in full.

Three Blood Drops, Two Worlds

Prosecution and defense experts use three blood drops, the roof geometry, a mannequin, reconstructed snow, and competing impact trajectories to argue incompatible accounts: Samuel was struck or pushed, or he fell and hit the shed on the way down.

A simulation can fit sparse observations while remaining non-identifying: different assumptions and test setups reproduce the evidence but imply opposite underlying mechanisms.

The experts openly acknowledge uncertainty, and forensic reconstruction is not model evaluation.

AI Risk families

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

AI safety concepts

This scenario is related to the following AI safety concepts.
  1. Capability Evals
  2. Calibration
  3. Red Teaming
  4. Distribution Shift