Gunnar Carlsson comments on Topological Data Analysis and Mechanistic Interpretability