gwern comments on Typicality and Asymmetrical Similarity