What is AI alignment?

Intro to AI Safety

Goal Misgeneralisation: Correct specs aren’t enough for Correct goals

Why AI alignment could be hard with modern deep learning?

Case study - (OpenAI) emergent tool use from multi-agent interaction