"You said we could come on the golf course, you didn't say anything about air horns, though!"
That is to say, you get permission/consent for one ostensible purpose, and then abuse that consent for something that is technically okay by the letter of the law, but goes way beyond what the user would agree to if they knew why you were asking for permission.
I don't trust that you'll keep this opt-in, as I have no reason to believe it.
--
Thanks.
I've disabled every permission messenger has until there is clarification here.
All the recognition happens on the device. Faces are detected, poses are estimated and frontalized, and features are extracted on the iphone. Then, the feature vector is compared to a small gallery of your friends' features living on the telephone. No upload.
That's what I hope happens.
I don't really see much a problem with the above potential strategy, but then again, I wouldn't want my recognition model living on my friends' phones. (One could imagine an attack that tries to extract the images used to train the recognition model given just the model blob...are there any private images there? sounds like a juicy paper!)
Or maybe the features are extracted but sent to Facebook, which returns a recognition result. I consider this less trustworthy because Facebook would learn the identity of everyone in any picture I take, even if I choose not to send the full picture to facebook.
Or: perhaps faces are detected, but then cropped and the image fragment containing the face is sent to Facebook.
Or more Orweillian still, perhaps the entire image is simply sent to Facebook and the recognition happens there.
If you were developing this feature, that is in my opinion a very interesting problem from technical point of view, then what were your thoughts about the possible social and moral implications of such feature?
For example, to give you an idea, did you think that this may outrage people? Or they may at least find it intrusive?
If you had any such thoughts, then how did you mitigate them?
Also is there some kind of program in Facebook that addresses these kind of problems?
I could not throw a rock at you. I am just interested about your personal experience.
There's a lot of openness, including around issues like this, at the company. There's weekly all hands meetings, where a developer could directly ask Mark about his opinion on a feature like this. The user experience, including creepiness, was almost certainly discussed. They may have misjudged in this case, but that's fallibility, not because it was ignored.