Atlas

The AI Safety Atlas is an open textbook on AI safety from the French Center for AI Safety (CeSIA) and partners. I’ve been working on two additions to it:

• A quiz companion: section quizzes and chapter reviews to check your understanding. It’s an unofficial proof of concept for CeSIA and covers chapter 1 so far.
• An audio read-along: the chapters’ audio is aligned to the text word by word, so the text highlights as it’s read and you can click any word to jump there.

[How you got involved, and why these two features]

The read-along works and is tested in the browser for chapter 1, on a branch of my fork of the Atlas site. The audio pipeline has processed all 8 chapters and 71 sections, and it notices when the source audio changes so it doesn’t redo work. The quiz’s questions are placeholders for now.

[What’s next: getting the read-along merged upstream, writing real quiz questions]