Could Apple get any training data out of this, given the homomorphic encryption?
You asked "If not [for training data], then why the default opt-in?" and I gave three alternative reasons.
a) Compensating for missing location metadata is a valid feature for a photo library. Peer competitor Google Photos also implements location estimation from landmarks.
b) The sizes of contemporary general vision models and the size of the vector database for matching potentially millions of landmarks suggest that this is not suitable for running on-device.
c) Apple's entire strategy is to do cloud computation without private data, so it stands to reason that they believe they're not using private data.
It’s one thing to argue that Apple is doing this for nefarious reasons, but to suggest that this is somehow the only conceivable option is a bit nuts.
Apple-scale companies have done far worse. It is hardly a "conspiracy" to allege Apple behaves like other companies of its ilk.
>It couldn’t just be that they think most people will want the feature?
Apple doesn't care what the end user thinks. "A lot of times, people don't know what they want until you show it to them" -Steve Jobs
>only conceivable option is a bit nuts.
I am merely suggesting it as the most likely option, not the only option.