Mechanistic interpretability

i.e. understanding what complex ML models are doing

• Chris Olah
• Catherine Olsson