This paper presents the first exponential separation for distribution independent learning of some natural problems in the interactive and non-interactive setting with differential privacy requirements. The reviewers felt this is a very good addition to the work on this topic.