Lee Sharkey comments on Efficient Dictionary Learning with Switch Sparse Autoencoders