Lee Sharkey comments on [Interim research report] Taking features out of superposition with sparse autoencoders