is apple cooked?
if you’ve ever used siri on your iphone, you know why people would think this…
and since apple hasn’t been developing massive foundation models like OpenAI or Google, many consider them far behind in the ai race.
but apple has actually been quietly making moves in ai, and most people haven’t noticed.
one of their main problems has been that they can’t run massive models like gpt-4 locally on your device.
and since one of their key selling points is security, using other models hosted on cloud infrastructure isn’t exactly ideal.
so what’s their solution?
create small but powerful models.
in june, they announced a set of new foundational models built specifically for apple silicon (their own hardware).
these on-device and server-based models will be available for app developers to integrate into their apps.
similar to their approach with the app store, apple is relying on developers to create the most value by building on top of these tools.
their on-device model competes with google’s gemma, which is also small but offers excellent performance.
however, apple’s model is less than half the size (3B parameters vs. 7B).
while all of this looks promising, it’s worth remembering that apple intelligence also looked promising… and didn’t really deliver on what was expected.
still, i’m writing this on my macbook, with my iphone in my pocket, airpods in my ears, and apple watch on my wrist
so I have to stay hopeful…