AI Alignment

Unit 1

AI and the years ahead

How and why might we build future AI systems?

Unit 2

What is AI alignment?

What do we need to do to ensure AI systems do what we want, and why is this difficult?

Unit 3

Reinforcement learning from human (or AI) feedback

Why do AI systems today mostly do what we want?

Unit 4

Scalable oversight

How might we scale human feedback for more powerful and complex models?

Unit 5

Robustness unlearning and control

Can we prevent oversight methods from being gamed?

Unit 6

Mechanistic interpretability

How might we understand what’s going on inside an AI model?

Unit 7

Technical governance approaches

How might we measure and mitigate the risks of deploying AI models?

Unit 8

Contributing to AI safety

How can you contribute?

Unit 9

Rapidly testing your project

Unit 10

Developing your project

Unit 11

Further developing your project

Unit 12

Building in public

Put your ideas on paper and share them with the world