Chris Olah - Looking Inside Neural Networks with Mechanistic Interpretability

Olah’s 2023 Alignment Workshop talk on reverse-engineering the internals of neural networks.