1. On device
2. Private cloud compute (aka Apple’s servers)
3. OpenAI (but only if you approve each time it’s used)
This isn’t really a new thing - you could always tell that some of Apple’s AI features depending on remote servers by disconnecting from the internet, using Siri and seeing how much of it stopped working.
Personally, I actually am somewhat encouraged that tech companies are subject to legal process (in the form of lawsuit discovery etc). Tech companies have paid billion-dollar settlements for allegedly breaking the laws in segments of their business that don’t generate billions in profits. Obviously there’s a lot of room for them to push the limits of the law, disagreement about what the law is, and there mays be ways for them to get away with things. But the fact that audits are happening, the code and logs can end up in court acts as some kind of constraint on their behavior.
“We’ll take the extraordinary step of making software images of every production build of PCC publicly available for security research. This promise, too, is an enforceable guarantee: user devices will be willing to send data only to PCC nodes that can cryptographically attest to running publicly listed software” [1].
If I can do AI on a computer in my pocket then why are we investing all of this money building AI Factories?