Does Mechanistic Interpretability Scale?

Evidence from Multiple Choice Capabilities in Chinchilla