Differential privacy is obviously worse for privacy than absolute privacy. But the latter is simply to restrictive and would preclude Apple from applying the be

Differential privacy is obviously worse for privacy than absolute privacy. But the latter is simply to restrictive and would preclude Apple from applying the benefits of machine learning to their users' benefit. I think, for example, what Apple is doing in health has some awesome potential. Also animated shit emojis! But "differential privacy" has a specific meaning that isn't just "sharing less data". It's trying to find provable methods to get the aggregate data needed for ML without exposing any single user's data. Kinda like the big data version of having one random soldier with a blank in his rifle in any firing squad.