AI Safety, Alignment & Governance
What could go wrong, and what to do about it
Research Lead · Safety, Alignment & Governance
About this course
A course for anyone who will deploy, regulate, procure or lead AI systems. It is technical enough to be honest and accessible enough to be useful: how models fail, how they are evaluated, what alignment research actually does, and how institutions in Pakistan and abroad are governing deployment.
Required for every diploma and executive cohort at the institute.
What you will be able to do
- Describe the main failure modes of deployed AI systems with real examples.
- Read an evaluation or model card critically.
- Design a governance process for an AI deployment in your organisation.
Curriculum
7 lessons · 1h 42m
- 01
How systems fail
- A taxonomy of failurePreview18 min
- Reading a model card14 min
- Module quiz · Failure modes6 min
- 02
Alignment and evaluation
- What alignment research does20 min
- Evaluation and red-teaming in practice16 min
- 03
Governance
- Governance that works in institutions18 min
- Course assessment10 min