Course-Wide Resources
This index collects resources linked from the Fields Institute course page, Professor Yevgeny Liokumovich's AI safety page, and a small set of papers matched to the weekly syllabus. Weekly folders contain shorter, topic-specific lists.
Suggested starting points
- ILIAD Intensive Curriculum — the course's suggested curriculum for foundational AI alignment research.
- Introduction to AI Safety, Ethics, and Society by Dan Hendrycks — the course's suggested introductory text. The chapters are readable online, and the site links to a downloadable PDF; the book itself is not stored in this repository.
- AI Safety for Mathematicians — mathematical introduction to the field, maintained by Jacob Tsimerman.
- Fields course lecture recordings — recordings for this course.
- AI Safety for Mathematicians video — linked from the instructor's page.
Related course notes
- Mathematics for AI Safety, Lionel Levine, Cornell University — lecture notes and references from a related mathematical course.
- AI Safety (CS 2881), Fall 2026, Boaz Barak, Harvard University — lectures on capabilities, training, interpretability, and alignment.
- AI Alignment, Roger Grosse, University of Toronto — linked from the Fields page; the page may require Notion access.
Related Safety Papers
See the independent relatedpapers folder for an annotated list of papers, grouped by course topic.
Research and community organizations
These are research, training, and community entry points listed on the instructor's page; their programs and availability can change.
- Mathematical AI Safety Institute (MAISI) — mathematical AI safety research.
- Fields Institute Centre for Mathematical AI — research themes, seminars, courses, and the joint Mathematics for AI Safety fellowship program.
- Toronto AI Safety Initiative (TAISI) — Toronto research training and student community.
- Trajectory Labs — Toronto AI safety workspace and research community.
- Fields–PrincInt postdoctoral fellowships in Mathematics for AI Safety — example research directions include interpretability, formal verification, and AI R&D automation risks.
Topic-specific readings
Use the weekly indexes for primary papers and focused reading lists:
- Weeks 01–02: capabilities and deep learning · Deep learning theory
- Weeks 03–04: training and adversarial attacks · Jailbreaking and adversarial attacks
- Weeks 05–06: representations and interpretability · Interpretability for control and alignment
- Weeks 07–08: interaction and oversight · Scalable oversight
- Weeks 09–10: guarantees and agency · Agency and decision theory