Short Course, AGI Safety

• Teach core concepts in AI alignment and AGI safety through recorded talks and exercises.
• Cover alignment problems from misaligned goals, including specification gaming and goal misgeneralization.
• Explore technical alignment approaches such as amplified oversight, interpretability, security, and safer design patterns.
• Examine institutional governance approaches to AI safety, including frontier safety practices and capability evaluations.

This course equips learners with foundational knowledge to contribute to the safe development of advanced AI systems, reducing risks from misaligned goals and promoting responsible AI governance.

• Interest in AI safety and alignment.
• Basic understanding of machine learning concepts.
• Ability to engage with technical and governance topics.

No compensation mentioned. This is a short course, not a job.

Google DeepMind is a world-leading AI research lab committed to developing artificial intelligence responsibly and ensuring its benefits are broadly shared.