Fellowships
We offer two parallel introductory fellowships in AI safety: alignment and governance.
The alignment track introduces the technical challenge of making AI systems reliably follow human intentions, while the governance track examines the role of policy, institutions, and global coordination to reduce AI risks.
Fellowships run weekly over dinner for 8 sessions in the form of paper discussions.
Applications are currently closed, and will reopen mid-March.
Curriculum developed by BlueDot Impact, adapted by TAISI.
Going in, I had some interest in AI safety but little idea how it shows up in real research or how someone technical like me could contribute. The curriculum and weekly discussions gave me a much clearer sense of the field, and I enjoyed the sushi.
Governance Fellowship
8 weeks
Topics include:
- Forecasting
- Overview of key actors
- Identifying levers for effective policy frameworks
- Governance at frontier labs
- Canada’s role in international cooperation
- Contributing to AI governance
Alignment Fellowship
8 weeks
Topics include:
- Intro to deep learning (first session only)
- Reinforcement learning from human feedback
- Scalable oversight
- Mechanistic interpretability
- Technical governance
- Contributing to technical AI safety