[another_kel]

I believe Google provides the weights, compute is apple owned

[pjmlp]

> Apple to use Google servers with Nvidia hardware for the new Siri

https://www.macworld.com/article/3156959/apple-to-use-google...

People have to stop thinking Apple is somehow different.

[coldtea]

From the link "Nvidia has its own “confidential computing” feature that encrypts data as it’s being processed, which will be used with other privacy and security measures to protect user data"

[speedgoose]

Do you know if it end to end encrypted or the keys are managed by Google?

[wmf]

I think the keys are directly from Nvidia and Intel.

[pjmlp]

Yes, and how does the data get there?

[wmf]

Over TLS?

[pjmlp]

To Google servers....

[wmf]

No, it's TLS to a "sealed" confidential VM that Google has no access to.

[kube-system]

From your link:

> Apple’s going to try to run as much of the new Siri as possible on-device

Anthropic and OpenAI don't have edge models.

[pjmlp]

Try is a magic word.

[kube-system]

It's not magic to those paying attention. The current model runs on device. And gemma e4b is already demonstrably capable on iPhone hardware, which is likely to be similar to what this is.

[kalleboo]

The new model only runs on device on iPhone 17 Pro or M3 Macs or newer. Anyone else is getting server compute.