Nathan Helm-Burger comments on Fluent dreaming for language models (AI interpretability method)