AI Safety, Alignment, and Governance

25 lessons in AI Fundamentals

• International Coordination and AI PolicySlides / Video
• Model Cards and Documentation for AccountabilitySlides / Video
• AI Standards and Risk FrameworksSlides / Video
• AI Governance Frameworks and PrinciplesSlides / Video
• Dual-Use and Misuse Risks of AISlides / Video
• Near-Term vs Long-Term AI RisksSlides / Video
• Corrigibility and Human ControlSlides / Video
• Hallucination and Truthfulness in LLMsSlides / Video
• Guardrails and Safety LayersSlides / Video
• Red Teaming AI SystemsSlides / Video
• Scalable Oversight and Supervision LimitsSlides / Video
• Distributional Shift and Robustness FailuresSlides / Video
• Instrumental Convergence and Goal MisgeneralizationSlides / Video
• Reinforcement Learning from Human FeedbackSlides / Video
• Value Alignment and Human PreferencesSlides / Video
• Specification Gaming and Reward HackingSlides / Video
• Outer vs Inner AlignmentSlides / Video
• The AI Alignment Problem DefinedSlides / Video
• What AI Safety Is and Why It MattersSlides / Video
• Compute GovernanceFlashcard
• AI Incident Reporting and MonitoringFlashcard
• Safety-Capability TradeoffsFlashcard
• Frontier Model Risks and EvaluationsFlashcard
• Watermarking and Provenance of AI ContentFlashcard
• Open vs Closed Model Release DecisionsFlashcard