Yeah also Prism - hello? Over 10 years ago we discovered that US can just enter any US based company’s server and read anything they want unless it’s directly encrypted but for these tools to function they have to decrypt data server side so LLMs can read the contents. Which means your data is not private in any way shape or form, not from Apple and not from US and not from anything in-between.
These claims by Apple are absolutely meaningless smoke for the ignorant who just follow tech buzzwords.
Since Private Cloud Compute needs to be able to access the data in the user’s request to allow a large foundation model to fulfill it, complete end-to-end encryption is not an option. Instead, the PCC compute node must have technical enforcement for the privacy of user data during processing, and must be incapable of retaining user data after its duty cycle is complete.
We designed Private Cloud Compute to make several guarantees about the way it handles user data:
A user’s device sends data to PCC for the sole, exclusive purpose of fulfilling the user’s inference request. PCC uses that data only to perform the operations requested by the user.
User data stays on the PCC nodes that are processing the request only until the response is returned. PCC deletes the user’s data after fulfilling the request, and no user data is retained in any form after the response is returned.
User data is never available to Apple — even to staff with administrative access to the production service or hardware.
And what is that supposed to guarantee? Who ever owns the hardware of unencrypted data owns it. There’s no way to pass tokens to LLM without unencrypting the content. Whatever path is made to obfuscation is fundamentally incapable of security.
There is on device AI in the Apple ecosystem. Many of the AI features that they announced will run locally (assuming hardware requirements are met). Things like Spatial Reframing will touch the cloud (via private compute) though. Other than that, Apple has an entire entry point for running AI close to the metal via MLX. It is kind of their entire angle at this point given their inability to create a competitive compelling AI product of their own. They appear to be taking on the role of “platform” once again.
Yes but these are nothing features compared to LLM. Samsung does on device picture moving nonsense too but in the grand scale it’s single digit % usage of AI if not below that. LLM is everything so the on-device tasks here are almost entirely meaningless.
While I loathe AI bullshit, Apple is at least prioritizing local, on-device AI and end-to-end encryption with their cloud AI services.
I’ll still be passing on any of this bullshit, but I appreciate that they tried to make a less problematic version.
What does end-to-end encryption even accomplish when you’re just feeding the information into an obscured, blackbox AI on the other end?
Like yes, I understand the importance of E2EE, I’m just making a point, it’s all rather ridiculous.
It’s not E2EE that matters most, but Apple’s Private Compute
Thank you, this is exactly true.
Most internet things are E2EE nowadays, but it matters not when the other end is AWS, Google, Cloudflare, or OpenAI.
Supposedly Apple claims to encrypt it at rest and in transit.
But data goes to the mothership anyway.
‘Bad actors’ can’t read your chatgpt conversations either, but OpenAI still does and can sell it.
Apple may better than Google, but I still don’t want my data there.
Yeah also Prism - hello? Over 10 years ago we discovered that US can just enter any US based company’s server and read anything they want unless it’s directly encrypted but for these tools to function they have to decrypt data server side so LLMs can read the contents. Which means your data is not private in any way shape or form, not from Apple and not from US and not from anything in-between.
These claims by Apple are absolutely meaningless smoke for the ignorant who just follow tech buzzwords.
Apple addressed that exact issue:
And what is that supposed to guarantee? Who ever owns the hardware of unencrypted data owns it. There’s no way to pass tokens to LLM without unencrypting the content. Whatever path is made to obfuscation is fundamentally incapable of security.
There’s no on device AI in Apple land - what are you talking about?
Also literally everything is end to end encrypted in this niche, that’s what s in https stands for.
There is on device AI in the Apple ecosystem. Many of the AI features that they announced will run locally (assuming hardware requirements are met). Things like Spatial Reframing will touch the cloud (via private compute) though. Other than that, Apple has an entire entry point for running AI close to the metal via MLX. It is kind of their entire angle at this point given their inability to create a competitive compelling AI product of their own. They appear to be taking on the role of “platform” once again.
Yes but these are nothing features compared to LLM. Samsung does on device picture moving nonsense too but in the grand scale it’s single digit % usage of AI if not below that. LLM is everything so the on-device tasks here are almost entirely meaningless.