Hacker News new | ask | show | jobs
by drawkward 527 days ago
Because apple could, just as easily, have auto opted-out everyone from this decision. I am not, by default, training data for AI or machine learning.
1 comments

Could Apple get any training data out of this, given the homomorphic encryption?
If not, then why the default opt-in?
Because it's a) a useful function for the app that b) can't practically be done entirely on-device, and c) they believe they're not sending anything private off-device.
None of that justifies signing me up for something without my permission.
That wasn't the question you asked.

You asked "If not [for training data], then why the default opt-in?" and I gave three alternative reasons.

None of which are sufficient justifications...
Apple turning a feature on by default can only possibly be a conspiracy to surreptitiously obtain training data from people’s photos? It couldn’t just be that they think most people will want the feature?

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 turning a feature on by default can only possibly be a conspiracy to surreptitiously obtain training data from people’s photos?

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.

I think it's technically impossible for Apple to get training data from this due to the encryption. In which case it's not a likely option at all.
The landmark tagging models are obviously trained on something. Apple still benefits from my involuntary participation.
The purpose of the homomorphic encryption is that they can compare similar landmarks without seeing the unencrypted image data on the server side.
Right, so I assume that they can’t build up a database of images to use for training future models. But I was hoping someone who understands homomorphic encryption and machine learning better than me could confirm this.