Actually, really liked the Apple Intelligence announcement. It must be a very exciting time at Apple as they layer AI on top of the entire OS. A few of the major themes.

Step 1 Multimodal I/O. Enable text/audio/image/video capability, both read and write. These are the native human APIs, so to speak.

Step 2 Agentic. Allow all parts of the OS and apps to inter-operate via “function calling”; kernel process LLM that can schedule and coordinate work across them given user queries.

Step 3 Frictionless. Fully integrate these features in a highly frictionless, fast, “always on”, and contextual way. No going around copy pasting information, prompt engineering, or etc. Adapt the UI accordingly.

Step 4 Initiative. Don’t perform a task given a prompt, anticipate the prompt, suggest, initiate.

Step 5 Delegation hierarchy. Move as much intelligence as you can on device (Apple Silicon very helpful and well-suited), but allow optional dispatch of work to cloud.

Step 6 Modularity. Allow the OS to access and support an entire and growing ecosystem of LLMs (e.g. ChatGPT announcement).

Step 7 Privacy. <3

We’re quickly heading into a world where you can open up your phone and just say stuff. It talks back and it knows you. And it just works. Super exciting and as a user, quite looking forward to it.

https://x.com/karpathy/status/1800242310116262150?s=46

  • RustmilianEnglish
    4•6 months ago

    However, to process more sophisticated requests, Apple Intelligence needs to be able to enlist help from larger, more complex models in the cloud. For these cloud requests to live up to the security and privacy guarantees that our users expect from our devices, the traditional cloud service security model isn’t a viable starting point. Instead, we need to bring our industry-leading device security model, for the first time ever, to the cloud.

    As stated above, Private cloud compute has nothing to do with the OS level AI itself. ರ⁠_⁠ರ That’s in the cloud not on device.

    While we’re publishing the binary images of every production PCC build, to further aid research we will periodically also publish a subset of the security-critical PCC source code.

    As stated here, it still has the same issue of not being 100% verifiable, they only publish a few code snippets they deam “security-critical”, it doesn’t allow us to verify the handling of user data.

    • It’s difficult to provide runtime transparency for AI in the cloud.
      Cloud AI services are opaque: providers do not typically specify details of the software stack they are using to run their services, and those details are often considered proprietary. Even if a cloud AI service relied only on open source software, which is inspectable by security researchers, there is no widely deployed way for a user device (or browser) to confirm that the service it’s connecting to is running an unmodified version of the software that it purports to run, or to detect that the software running on the service has changed.

    Adding to what it says here, if the on device AI is compromised in anyway, be it from an attacker or Apple themselves then PCC is rendered irrelevant regardless if PCC were open source or not.

    Additionally, I’ll raise the issue that this entire blog is nothing but just that a blog, nothing stated here is legally binding, so any claims of how they handled user data is irrelevant and can easily be dismissed as marketing.

    • @Z4rKOPEnglish
      -3•6 months ago
      1. Security / privacy on device: Don’t use devices / OS you don’t trust. I don’t see what difference on-device AI have at all here. If you don’t trust your device / OS then no functionality or data is safe.
      2. Security / privacy in the cloud: The take here is that Apples proposed implementation is better than 99% of every cloud service out there. AI or not isn’t really part of it. If you already don’t trust Apple then this is moot. Don’t use cloud services from providers you don’t trust.

      Security and privacy in 2024 is unfortunately about trust, not technology, unless you are able to isolate yourself or design and produce all the chips you use yourself.

      • RustmilianEnglish
        5•6 months ago

        Yeah and apple is completely untrustworthy like any other corporation, my point exactly. Idk about you, but I’ll stick to what I can verify the security & privacy of for myself, e.g. Ollama, GrapheneOS, Linux, Coreboot, Libreboot/Canoeboot, etc.

        • @Z4rKOPEnglish
          0•6 months ago

          Ok, I just don’t see the relevance to this post then. Sure, you’re fine to rant about Apple in any thread you want to, it’s just not particularly relevant to AI, which was the technology in question here.

          I hear good things about GrapheneOS but just stay away from it because of all the stranger. I love Olan’s.

          • RustmilianEnglish
            4•6 months ago

            We’re discussing Apple’s implementation of an OS level AI, it’s entirely relevant.

            GrapheneOS has technical merit and is completely open source, infact many of the security improvements to Android/AOSP are from GrapheneOS.

            I love Olan’s.

            Who?

            • @Z4rKOPEnglish
              2•6 months ago

              Lol thank you autocorrect. Ollama.