Tag: opacity

AI Systems

On Trusting Systems You Cannot Fully Inspect

Deploying opaque AI systems creates tension between utility and epistemic responsibility. Current interpretability explains roughly 30-40% of model behavior.