Results for "misalignment"
Value Misalignment
AdvancedModel optimizes objectives misaligned with human values.
Value misalignment is like a situation where a robot is told to make people happy but ends up doing things that actually upset them. Imagine a robot programmed to serve ice cream but only gives out flavors that people don’t like. In AI, this happens when the system's goals don’t match what humans...
Value Misalignment
Advanced
Model optimizes objectives misaligned with human values.
Alignment Problem
Advanced
Ensuring AI systems pursue intended human goals.
Outer Alignment
Advanced
Correctly specifying goals.
Alignment Tax
Advanced
Tradeoff between safety and performance.
Deceptive Alignment
Advanced
Model behaves well during training but not deployment.
Existential Risk
Advanced
Risk threatening humanity’s survival.