Safety and Reliability¶
Risk properties of AI systems, independence requirements, common-cause failure, and the challenges of verifying stochastic systems.
Risk Properties¶
- non-determinism-and-reproducibility — Stochastic output, V&V gap, model version instability
- degradation-characteristics — Gradual, catastrophic, and silent degradation patterns
- opacity-and-explainability — Double opacity, provider filtering, auditability barriers
- deployment-local-vs-cloud — Local vs cloud deployment, open-weight models, configuration management
Independence and Common-Cause Failure¶
- epistemic-independence — Independence of reasoning and evidence in multi-agent systems
- monoculture-collapse — Common-cause failure from shared model weights and training data
Verification and Validation¶
Pages to be added during subsequent ingestion.