That is precisely my point. Differential privacy would NOT make any difference, and I was pointing the many folks who are working on it to the much simpler issues that are in fact being encountered in the field. This past IEEE S&P had quite a few theoretical privacy talks.