This is a link post for Michael Nielsen's "If correlation doesn’t imply causation, then what does?" (2012).

I want to highlight the post for a few reasons:

(1) it is a well-written introduction by an experienced science communicator — Michael is an author of the most famous book on quantum computing;

(2) causal inference is an essential tool for understanding the world;

(3) two recent AI safety papers use causal influence diagrams to (a) understand agent incentives [arXiv, Medium] and (b) to provide a new perspective on some problems in AGI safety [arXiv].

New Comment