The Iliad Intensive Course Materials
The Iliad Intensive is a month-long, in-person AI alignment course offered every other month, designed for individuals with strong technical backgrounds in mathematics, physics, or theoretical computer science. The course materials, developed by around 20 contributors, include lecture notes, mathematical exercises, coding problems, and self-contained modules on advanced topics like singular learning theory and data attribution. These materials are being released to promote common knowledge, invite feedback, and support independent learning, with plans for ongoing updates and a dedicated website.
- ▪The Iliad Intensive is a full-time, month-long course focused on AI alignment, held in-person every other month.
- ▪The course targets individuals with strong backgrounds in mathematics, physics, or theoretical computer science and includes technical content such as mathematical exercises, coding problems, and lecture notes.
- ▪The materials are structured into clusters and modules covering topics like AI alignment frameworks, alignment in practice across model development phases, and reward learning theory, with plans for continuous improvement and future release
- ▪A prerequisite module outlines necessary technical knowledge including deep learning, linear algebra, calculus, probability, and information theory, along with foundational readings on AI safety and its importance.
- ▪The release aims to build common knowledge, enable independent study, and collect feedback, with updates to be shared as significantly revised versions of the materials are developed.
Opening excerpt (first ~120 words) tap to expand
We are releasing the course materials of the Iliad Intensive, a new month-long and full-time AI Alignment course that runs in-person every second month. The course targets people with strong backgrounds in mathematics, physics, or theoretical computer science, and the materials reflect that: they include mathematical exercises with solutions, self-contained lecture notes on topics like singular learning theory and data attribution, and coding problems, at a depth that is unmatched for many of the topics we cover.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Lesswrong.
Discussion
0 commentsMore from Lesswrong
