undefined | Better HN

0 pointshu31y ago0 comments

Am I understanding correctly that some AI will run on Apple servers? So not completely offline AI.

If so that's somewhat disappointing given how much AI power Apple hardware packs.

> Apple sets a new standard for privacy in AI, with the ability to flex and scale computational capacity between on-device processing and larger, server-based models that run on dedicated Apple silicon servers

0 comments

onesociety20221y ago

Maybe that is actually good for iPhone buyers? Otherwise every 2 years, they will claim a bunch of new AI features will not work on your older device. But if they can delegate requests to a server, older devices can continue receiving newer AI features from future iOS releases (they will just be slower than the newer iPhones which will run them locally).

dylan6041y ago

The interesting thing I took from that is they are making servers with M-series chips. Maybe they're just rack mounted Mac Minis? But if Apple decides that crap, maybe it'll get them to a point where they decided to make a proper rack mountable form factor available??? <genmojiTechPrayingInFrontOfRack>

Almondsetat1y ago

I think it's a little disingenuous to think on-device accelerators on a mobile phone would be able to do literally any AI task without help

ducadveritatem1y ago

Based on their presentations yesterday and accompanying written materials they have a pretty capable ~3B parameter Foundational LLM running fully locally on the devices which is the first line. https://machinelearning.apple.com/research/introducing-apple...

j / k navigate · click thread line to collapse