Resources: Mechanistic interpretability

Exercises: Mechanistic interpretability

Resources: Rapidly testing your project

Exercises: Rapidly testing your project

Resources: Building in public

Exercises: Building in public

Resources: Developing your project

Exercises: Developing your project

Resources: Technical governance approaches

Exercises: Technical governance approaches

Resources: AI and the years ahead

Exercises: AI and the years ahead

Resources: Contributing to AI safety

Exercises: Contributing to AI safety

Resources: What is AI alignment?

Exercises: What is AI alignment?

Resources: Further developing your project

Exercises: Further developing your project

Resources: Scalable oversight

Exercises: Scalable oversight

Resources: Robustness unlearning and control

Exercises: Robustness unlearning and control

Resources: Reinforcement learning from human (or AI) feedback

Exercises: Reinforcement learning from human (or AI) feedback

**This course has been replaced by our new [Technical AI Safety course](https://bluedot.org/courses/technical-ai-safety).**

Quickly validating your project idea

We're excited for you to operate like an entrepreneur in approaching your project:

- pick an initial idea (you did this last unit!)
- run an experiment to test your idea
- use the results to improve the idea, or throw the idea out the window and find a new one

Large projects often fail because they start with an awesome-sounding idea, spend months building something, and only then discover that it's fundamentally flawed.

You should therefore start by validating the feasibility and usefulness of your project idea in the quickest way possible.

In the unit, you'll present your results to your group, and flesh out your action plan for the next weeks.


Describe the validation work or cheap test you've conducted, what the results were, and the implication for the next few weeks of your project.


Your cheap test

AI Alignment

Rapidly testing your project

How and why might we build future AI systems?

AI and the years ahead

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

What is AI alignment?

Why do AI systems today mostly do what we want?

Reinforcement learning from human (or AI) feedback

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

Scalable oversight

Can we prevent oversight methods from being gamed?

Robustness unlearning and control

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

Mechanistic interpretability

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

Technical governance approaches

Contributing to AI safety

Developing your project

Further developing your project

Put your ideas on paper and share them with the world