Apple Plans Paid iCloud+ Upgrades for Heavy AI Users

[michaeljk]

Genmoji and flight times are trivial examples of what Apple is talking about. In 2-3 years, as apps fade into the background (they will no longer be tolerated by users who want the OS to handle complex tasks without having to switch to different apps for specific actions), accelerated by competition from OpenAI, Anthropic and others that will release their own OSes and hardware, Apple will begin to lose substantial revenue from the App store. Apple users will no longer want to buy Apps, own Apps, use Apps. Apps will become pointless, obstructive, dumb. This is good for Apple, not so good for independent developers (unless they can somehow kludge themselves as add-ons to the Apple OS with actual useful AI features, which Apple will likely thwart). Apple is getting very busy to get ready to Sherlock all the other apps with its AI. To offset its losses from the App Store and developer community, it will leverage the one thing it still has: a massive user base wedded to its ecosystem. Combine that with its pretty impeccable record on user privacy and security, it will tout this as a reason to use its AI tools, either on device (with rapidly developing neural networks built into its own machines, using its own hardware) or in the cloud, with "Private Cloud Compute." This is genius. It might also mean they will be less reliant on AWS, Azure, or Google to provide their cloud computing needs.

I never let my AI tools anywhere near any of my files. I do not allow any of my AI tools to store any of my projects or conversations in their cloud storage. I don't really like not having much control over how Anthropic or Perplexity (and its constituents) might be using my questions, data, interactions without my knowledge and permission. I know others will think this is naive, but I will disagree: I trust Apple with my privacy. You know why? Apple doesn't make money on my data. They might strategize how to sell me more stuff based on my data, but they make zero dollars selling my data to anyone else. Who else can say that? Also, Apple has made huge efforts to sell itself as not just a hardware company, or a software company. They've done a good job selling themselves as a privacy, security, and data protection company. This will become even more important going forward. They have the hardware and the OS that allows users right now to do all kinds of things with AI models entirely on device, which is apparently why the Mac mini (and maybe Mac Studio?) are outselling Apple's capacity to make them. Brilliant. If they can get to where they need to be before their competitors do (new competitors in the AI space).

 

[capamac]

The LLMs will hallucinate your flight time, causing you to miss your expensive flight.

If we ever get AGI, Apple and every software company will probably become irrelevant. Whatever software moat Apple has will vanish. Maybe Apple can still make money selling hardware alone.

But LLMs are not AGI. LLMs are just very silly fake "AI" toys.

 

[deconstruct60]

If you code, then Siri AI is not for you--presently, anyway. I am sure we will be able to code macOS and iOS apps using Siri AI in the coming years. Or at least it is on the table for discussion

'Code' in general. But doubtful there is a huge gap in composing Shortcuts with Apple PCC and one of the other models primarily trained to do other languages and application types.

Use Apple Intelligence in Shortcuts on iPhone

In Shortcuts on iPhone, use a model to handle simple or complex requests in your shortcuts.

support.apple.com

 

[Padre_Pio]

Genmoji and flight times are trivial examples of what Apple is talking about. In 2-3 years, as apps fade into the background (they will no longer be tolerated by users who want the OS to handle complex tasks without having to switch to different apps for specific actions), accelerated by competition from OpenAI, Anthropic and others that will release their own OSes and hardware, Apple will begin to lose substantial revenue from the App store. Apple users will no longer want to buy Apps, own Apps, use Apps. Apps will become pointless, obstructive, dumb. This is good for Apple, not so good for independent developers (unless they can somehow kludge themselves as add-ons to the Apple OS with actual useful AI features, which Apple will likely thwart). Apple is getting very busy to get ready to Sherlock all the other apps with its AI. To offset its losses from the App Store and developer community, it will leverage the one thing it still has: a massive user base wedded to its ecosystem. Combine that with its pretty impeccable record on user privacy and security, it will tout this as a reason to use its AI tools, either on device (with rapidly developing neural networks built into its own machines, using its own hardware) or in the cloud, with "Private Cloud Compute." This is genius. It might also mean they will be less reliant on AWS, Azure, or Google to provide their cloud computing needs.

I never let my AI tools anywhere near any of my files. I do not allow any of my AI tools to store any of my projects or conversations in their cloud storage. I don't really like not having much control over how Anthropic or Perplexity (and its constituents) might be using my questions, data, interactions without my knowledge and permission. I know others will think this is naive, but I will disagree: I trust Apple with my privacy. You know why? Apple doesn't make money on my data. They might strategize how to sell me more stuff based on my data, but they make zero dollars selling my data to anyone else. Who else can say that? Also, Apple has made huge efforts to sell itself as not just a hardware company, or a software company. They've done a good job selling themselves as a privacy, security, and data protection company. This will become even more important going forward. They have the hardware and the OS that allows users right now to do all kinds of things with AI models entirely on device, which is apparently why the Mac mini (and maybe Mac Studio?) are outselling Apple's capacity to make them. Brilliant. If they can get to where they need to be before their competitors do (new competitors in the AI space).

But....does Apple want to throw itself into the group of other AI models and deal with those headaches and regulatory issues? They don't even have a deal in place for the EU.

 

[deconstruct60]

Genmoji and flight times are trivial examples of what Apple is talking about. In 2-3 years, as apps fade into the background (they will no longer be tolerated by users who want the OS to handle complex tasks without having to switch to different apps for specific actions), accelerated by competition from OpenAI, Anthropic and others that will release their own OSes and hardware,

The whole 'apps will go away' thing seems likely overblown. At least for folks who don't want to get on fastpath to being dependent up agents for all critical thinking. AI is wrong often enough that want to be able to audit whether they messed up or not. If surrender all skills to the agents as to what the apps do .... how are you going to audit them?

Apps that are relatively very thin veneers over web services (e.g., web service vendor wanted to avoid doing it ina plain browser webpage because avoiding ads/browser security, etc) ... maybe. But deeply substantive apps. Probably not.

As far as hardware from the hyper cloud focused AI vendors. ... chuckle believe it when I see it. I have no doubt OpenAI is currently blowing gobs of money on this. But whether it works or not is a different story. So far it mainly looks to be a mechanism of sucking your privacy dry. What "OS" it runs doesn't particularly matter, because it may not be user purposed. Chatbot-crackheads will suck them down, but I doubt they really add much that an app on iOS/iPadOS wouldn't already do.

Again these vendors are cloud focus and their device is like cloud focused. There is no OS in part because there would be no apps there. In part also to protect because their AI generated apps have no IP protection , so need to 'hide' them in the cloud so get 'trade secret' protection. That is a weakness, at least as much as a strength.

They have the hardware and the OS that allows users right now to do all kinds of things with AI models entirely on device, which is apparently why the Mac mini (and maybe Mac Studio?) are outselling Apple's capacity to make them. Brilliant. If they can get to where they need to be before their competitors do (new competitors in the AI space).

The OpenClow and the other mania variants don't particularly run the larger LLMs locally. There may be something for the local coordinating/orchestrating bot to do, but the delegations can be off the box. ( OpenClaw has a setting on local or clould location for the LLM). 24-64GB seems about range to get more complicated things done and that is at the edge of the Mini. But can go Mini and 'punt' the LLM aspect into the cloud. (perhaps running the apps/tools locally but 'what to do' is coming back from out in the cloud. )

 

[michaeljk]

The whole 'apps will go away' thing seems likely overblown. At least for folks who don't want to get on fastpath to being dependent up agents for all critical thinking. AI is wrong often enough that want to be able to audit whether they messed up or not. If surrender all skills to the agents as to what the apps do .... how are you going to audit them?

Good point, for now. I am not suggesting this is going to happen all at once, or even in the very near future. That's why I used the term "fade." iOS 27 will be Apple's first real attempt at this, and from the looks of it, what we may end up finding is a limited set of "natural language" capabilities to create on the fly shortcuts. I mean, the ease with which we already do this in very limited fashion speaks to how much we'd all like to be able to do this much more simply. I just told my home pod, "Hey Siri, play drone music" and that's what it is doing. I assume it is using Apple Music, but don't really know, don't really care. Apple knows this is where its users want to go. Apps are clunky. I didn't have to pick up my iPhone, go into the "Music" app, type in "Drone," find something I like, then hit play or shuffle. Yes, this is a simple example, and has existed for a few years, but we are on the cusp of much more complex tasks. Also, this idea that AI is static doesn't make any sense. The number of errors the models made even one year ago don't compare to now. So, what's it going to be like in 5 years.

Also, I want computers to begin making inferences more often. I don't always want to have to tell a computer exactly what I want, because it is a hassle, and because I myself might not know exactly what I want. If I can trust an AI's privacy, I absolutely want it to store and use all of my previous choices to make inferences about what I want in the present moment. Of course it will sometimes get it wrong, but this doesn't mean it's a mistake. I get my own inferential tendencies wrong all the time. In human speak, we call this changing our mind.

 

[capamac]

I am no "AI" expert but I listened to a youtube chat last month that argued the "AI" systems are basically using cheap hacks to try and reduce errors. There's no actual increase in "intelligence." They're just running their initial response through a second "AI" query in an attempt to filter out mistakes. But this costs more computationally, and "AI" is already extremely expensive and priced far below actual cost. The author also argued that specific problems such as counting the number of "R" characters in "strawberry" have been manually fixed. The "AI" didn't "fix" the strawberry problem -- a human programmed the "fix" into the system. The whole "AI" thing just strikes me as b.s.

 

[realLucaR]

Name five people that unironically use image playground. I'll wait.

 

[deconstruct60]

Name five people that unironically use image playground. I'll wait.

Chronically (as opposed to un-ironically)... don't have to get most folks addicted to make money.

if have 500M Messages users and only 0.1% of them sprinkle AI gen images ( Genmojii , deep AI altering photos , etc ) then that is 500K users. If upgrade 250K of them to $10/month more that is about the is $30M more a year. )
750M users same percentages it is $45M. Take that baseline and go to 1% that is $450M.

I suspect $450M could pay off more than just a couple of Private Cloud Compute node costs in the span of a year.

Stickers , genmojii, emojiis .... how many people use zero of these. Slapping pictures in messages is a feature that more thana few folks have bought into.

 

[eleuthero]

Chronically (as opposed to un-ironically)... don't have to get most folks addicted to make money.

if have 500M Messages users and only 0.1% of them sprinkle AI gen images ( Genmojii , deep AI altering photos , etc ) then that is 500K users. If upgrade 250K of them to $10/month more that is about the is $30M more a year. )
750M users same percentages it is $45M. Take that baseline and go to 1% that is $450M.

I suspect $450M could pay off more than just a couple of Private Cloud Compute node costs in the span of a year.

Stickers , genmojii, emojiis .... how many people use zero of these. Slapping pictures in messages is a feature that more thana few folks have bought into.

While I think you're probably right, I am also a counter example for you. I make use of four emojis—the smiley face, the frowny face, the winky face (all three from when it was ASCII only and I just kept going), and the thumbs up, all of which were created by artists (I think) prior to modern AI gen. Of course, I also never got into the whole sending of animated gifs thing either.