Alignment Problem

Advanced

Ensuring AI systems pursue intended human goals.

Full Definition

Ensuring AI systems pursue intended human goals.

Keywords

intent mismatch

Domains

AI Safety & Alignment

Related Terms

Specification Gaming related to

Model exploits poorly specified objectives.

 Reward Hacking related to

Maximizing reward without fulfilling real goal.

 Instrumental Convergence related to

Tendency for agents to pursue resources regardless of final goal.

 Value Misalignment related to

Model optimizes objectives misaligned with human values.

Concept Map

See how Alignment Problem connects to other concepts.

Open Knowledge Graph