Results for "Shapley values"

Existential risk from AI systems.

Tradeoff between safety and performance.

Isolating AI systems.

Signals indicating dangerous behavior.

Tendency to gain control/resources.

Intelligence and goals are independent.

Research ensuring AI remains safe.

Designing AI to cooperate with humans and each other.

A formal privacy framework ensuring outputs do not reveal much about any single individual’s data contribution.

Measures a model’s ability to fit random noise; used to bound generalization error.

Learning only from current policy’s data.

Risk threatening humanity’s survival.

Inferring the agent’s internal state from noisy sensor data.