3 ms·

This is a slight aside, but could someone more familiar with the field than me explain what the Neural Engine is for? Does any software currently use it, or is

This is a slight aside, but could someone more familiar with the field than me explain what the Neural Engine is for? Does any software currently use it, or is it something that Apple are trying to push? Is it designed to build ML models, or to apply them to data? And perhaps more basically, what for? How much machine learning is done on user machines, as opposed to renting some cloud time? What are they anticipating will be done? I ask these questions in genuine curiousity, I assume I'm missing something but this just seems like a rather wild divergence from what (very little) I knew of the ML field.

marta_morena_28 6y ago
"Core ML optimizes on-device performance by leveraging the CPU, GPU, and Neural Engine while minimizing its memory footprint and power consumption." https://developer.apple.com/documentation/coreml https://developer.apple.com/documentation/coreml Apple has always been good at running ML on device, as opposed to Google's approach of sending everything to the cloud, then mining all your data, selling it, tracking it, etc. One example is iPhotos, which implements feature detection on-device, while Google implements it in the cloud. https://www.imore.com/no-apples-machine-learning-engine-cant-surface-your-iphones-secrets https://www.imore.com/no-apples-machine-learning-engine-cant... Or for Google: https://www.theringer.com/2017/5/25/16043842/google-photos-data-collection-e8578b3256e0 https://www.theringer.com/2017/5/25/16043842/google-photos-d... (Reading this makes me wanna puke)
CountSessine 6y ago
It’s accessible through the CoreML framework. I know nothing about this and very little about ML in general, but I found an interesting GitHub page written by a consultant who seems to know a fair amount about it. https://github.com/hollance/neural-engine https://github.com/hollance/neural-engine
saagarjha 6y ago
It does matrix multiplication and stuff that is useful for neural networks.

marta_morena_28 6y ago

"Core ML optimizes on-device performance by leveraging the CPU, GPU, and Neural Engine while minimizing its memory footprint and power consumption." https://developer.apple.com/documentation/coreml https://developer.apple.com/documentation/coreml Apple has always been good at running ML on device, as opposed to Google's approach of sending everything to the cloud, then mining all your data, selling it, tracking it, etc. One example is iPhotos, which implements feature detection on-device, while Google implements it in the cloud. https://www.imore.com/no-apples-machine-learning-engine-cant-surface-your-iphones-secrets https://www.imore.com/no-apples-machine-learning-engine-cant... Or for Google: https://www.theringer.com/2017/5/25/16043842/google-photos-data-collection-e8578b3256e0 https://www.theringer.com/2017/5/25/16043842/google-photos-d... (Reading this makes me wanna puke)

CountSessine 6y ago

It’s accessible through the CoreML framework. I know nothing about this and very little about ML in general, but I found an interesting GitHub page written by a consultant who seems to know a fair amount about it. https://github.com/hollance/neural-engine https://github.com/hollance/neural-engine

saagarjha 6y ago

It does matrix multiplication and stuff that is useful for neural networks.