If you followed anything in Apple's WWDC2020 you'll have seen that iOS14 was attempting to secure proprietary app data. CoreML Model encryption was one notable announcement.
Apple has also updated, last year, their DeviceCheck API to include an App Attest Service which lets developers check if an installation of an App is compromised before downloading data (e.g. AI model) which is at risk of being stolen. I believe it requires the T2 secure enclave.
For an app developer this is great because the only way to protect your valuable AI ML models right now is to offer them as a cloud based service.
CoreML Model encryption is meant to encourage downloading of AI models onto the device and run without the need for a network connection which may even provide speed gains.
Apple device owners will benefit from T2 being a requirement (Mac OS too now) as they will benefit from faster, and offline, AI applications becoming available on both iOS and Mac OS.
I'm surprised at the iPhone X omission from some of the on device AI applications like voice dictation but not concerned.
Too many people though are jumping onto the "this is a marketing move to force upgrade".
It is more likely to be a hardware issue.
The X/8's A11 was the first of the Bionic series including neural engine.
If I was to hazard a guess I'd say that either the A11 isn't up to the task required for these features or that the T2 chip in the X/8 iPhones isn't secure enough for Apple to be happy to allow downloading of their ML models onto those devices. That and memory requirements (for CPU, GPU or now even NPU) . That always hold back older models from getting the new stuff.
Apple has also updated, last year, their DeviceCheck API to include an App Attest Service which lets developers check if an installation of an App is compromised before downloading data (e.g. AI model) which is at risk of being stolen. I believe it requires the T2 secure enclave.
For an app developer this is great because the only way to protect your valuable AI ML models right now is to offer them as a cloud based service.
CoreML Model encryption is meant to encourage downloading of AI models onto the device and run without the need for a network connection which may even provide speed gains.
Apple device owners will benefit from T2 being a requirement (Mac OS too now) as they will benefit from faster, and offline, AI applications becoming available on both iOS and Mac OS.
I'm surprised at the iPhone X omission from some of the on device AI applications like voice dictation but not concerned.
Too many people though are jumping onto the "this is a marketing move to force upgrade".
It is more likely to be a hardware issue.
The X/8's A11 was the first of the Bionic series including neural engine.
If I was to hazard a guess I'd say that either the A11 isn't up to the task required for these features or that the T2 chip in the X/8 iPhones isn't secure enough for Apple to be happy to allow downloading of their ML models onto those devices. That and memory requirements (for CPU, GPU or now even NPU) . That always hold back older models from getting the new stuff.