Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple is great at physical consumer products but I feel like the etherealness of AI has been really difficult for them grasp and thus adopt into their products in a meaningful way.

We've had Siri since 2011 and not to harp on it too much but it hasn't changed all that much, evolutionary steps but it feels very "hard coded" in that you have to ask it things Apple has specifically built what I'd call scenarios for.

What people really want (I think) is Siri to act like the LLM's do, specifically the ChatGPT app on iPhone which has a human-like voice and conversational context. If you can talk to Siri from anywhere on the system and ask it to do a task no matter how complicated (opening the browser, going to a specific site, searching for a product, telling you the price and availability etc) then it will meet and eclipse some of these other AI agents available.

Whether Apple can pull that off I don't know, seems like they are concerned about privacy and how to deliver on AI on-device but they've been limiting the amount of RAM they provide on their entire device lineup for a decade so now they're caught with their pants down trying to run large models on-device. Thus they turn to the cloud, which is not as private.

It's a real conundrum. I do agree with Tim that Apple has rarely been first but when it comes to this AI stuff they actually were one of the first companies to implement AI into their products but they have done so very poorly. They've really not been waiting for an oppertunity they jumped in and everyone has collectively yawned. I hope they can turn it around and these words from them are aligned with their actions.
 
Neither do you.
You're just retconning everything to fit how it goes over time, same as Apple.

It's pretty clear what Apple's AI strategy is. And that's investing $500 Billion to manufacture its own proprietary AI servers with a focus on user privacy/security in its Houston manufacturing facility. And then deploying them around the country (and likely beyond).
 
Cook pointed out that Apple has dominated several markets even when the company wasn't first to the technology. "We've rarely been first.
Plot twist:
They pretty much were the first with Siri:

"Not only did it understand the words I said. It understands the meaning." (Scott Forstall, 2011)
 
Last edited:
He still thinks this is just hype and that he can fix it by throwing money at the problem, as if it is simply a supply chain issue.

Companies are spending billions on AI hardware to secure the best chips for their servers. Meanwhile, he thinks M2s will do the job.

Competitors are poaching top talent and spending hundreds of millions, much of it coming straight from his own company.

Your rivals are surpassing your market capitalisation, which seems to be the only metric that matters to you, since shareholder value clearly takes priority over customer satisfaction.

And worst of all, you are losing the trust of your most loyal customers by announcing vapourware.

Yet this is the speech you deliver at your town hall meeting? Something is seriously broken.
 
Totally agree Francisco

I'll say it again -- it's refreshing to see how many here aren't buying the Apple BS narratives anymore.


Screenshot 2025-08-01 at 16.52.12 Redacted.jpeg
 
The Apple AI models all there, we can see ways to use it in the beta in Shortcuts. They will definitely get this to where it needs to be.
Wow, I didn't even realize that. I dug in deeper and I actually created a Shortcut where I can give it a prompt, and it uses Apple's on-device models to respond. I tried to put it to the test and asked what Sandy Koufax's career WHIP was. And it responded correctly, and quickly. All on device, in airplane mode. I'm actually quite impressed.
 
Time for what?
To address his employees and acknowledge reality.

If I'm reading your other post correctly I think you're way off base on Apple having a grand strategy for AI that is baking behind the scenes. Private Cloud Compute has a great implementation but Apple can't wait a decade and get a privacy centric version out when the sheer utility means people will make that trade willingly.

Google has about 500 million users of Gemini, and ChatGPT is on track for 1 Billion users by the end of the year. Neither of those companies are great with privacy; OpenAI isn't even deleting anything unless you have a ZDR contract with them due to the NYT lawsuit.

There is not a market advantage here until ads roll out and some of the providers (e..g Meta, xAI) have enough capital to stave that off for the rest of this decade if people really revolt against it when it occurs.

There's a paradigm shift happening with the entire internet for many reasons, a lot of them pretty nefarious. RealID is being required virtually simultaneously worldwide with many countries and many states in the US passing laws for various sites and services. That's going to accelerate, unfortunately.

I care deeply about privacy and especially tracking but I make tradeoffs to use frontier models because they provide enormous utility for specific tasks. I gate what I put in there and still use them fairly often. I'm for sure in the less than 1% of the billion+ using AI that cares about that.

Apple needs a quality advantage, privacy alone won't be enough. Nvidia is also probably 2 years ahead of them on hardware, even with the stuff we haven't seen publicly. I'd bet my house on that.

I want Apple to catch-up and surpass the others in user experience, but they need to retain researcher talent and grow it – acquiring Anthropic and keeping it as a subsidiary as I've posted elsewhere would be my choice. I think perplexity is a mistake but it might be a hedge in case they lose the Google deal, so they can have the appearance of killing two birds with one stone, getting search and AI. I just don't think perplexity has a unique advantage or novel research which is critical, particularly with world models which will determine who stays relevant into the 2030s.

I don't ascribe to the theory that the first to world models will dominate forever, but I do think it's extremely likely the research will be proprietary so multiple independent efforts are critical both for society at large and to make sure there isn't some patent nonsense down the road because the equivalent of licensing something as powerful as the internet is probably not going to happen. Meta (who probably has the largest lead right now internally) is already starting to seed the idea that they won't, in fact, keep open sourcing key research – it's even in Zuckerberg's post from a couple days ago.

Apple may offer the platform that we interface with but if they want to own the whole stack they have a ton of work to do. I think their efforts to date have been mostly misfires outside of their good research on small quantized models which are useful but are not the type of thing that will be "bigger than the internet" as Tim put it. That requires enormous levels of investment in R&D and the management commitment to keep both the talent and hardware to support them maximally funded.
 
Last edited:
Apple may offer the platform that we interface with but if they want to own the whole stack they have a ton of work to do. I think their efforts to date have been mostly misfires.

I dunno... If Apple with their $500 Billion investment will soon be manufacturing their own AI servers this year, I would guess any Srouji team AI chips are pretty far along and they have already been testing first silicon. Pure speculation of course, but makes sense. And Apple would certainly not be making any of that public at this point.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.