Apple could have done this too for devices < 8GM RAM but they didn't. "privacy" bla bla they already have everything on iCloud and they already have an uncheck button for AI.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Report: Apple's AI Strategy Could Finally Pay Off in 2026
- Thread starter MacRumors
- Start date
- Sort by reaction score
Just like they license many other software from other companies, or do you think Apple develops every single thing they use? They are at the mercy of their competitors as much as they are at the mercy of Samsung that makes their iPhone screens.
Profit potential there is, but it’s almost certainly not in the consumer market and not enough for every company that has invested dozens and even hundreds of billions in it. In the end it’s probably going to be Google and ChatGPT standing with Microsoft limping along as always and a few smaller companies like Midjourney serving niche use cases. Apple has nothing to look for here unless they want to make a push for enterprise. They might be better off using task specific models as they’ve been doing always.
Are your experts paid by AI companies? Not withstanding there are also many experts, also paid by AI companies themselves that are now doubtful of whether LLMs can achieve AGI. It was always an uncertain thing as there is more to agency and rationality than language and algorithmic processing. But if you are raising money from investors you cannot be transparent about the uncertainty of the whole thing. Also if you dig a little deeper you’ll find this is not the first time tech companies thought it was close to true AI. Maybe this is the one, but as time passes it seems more likely that we’re still not quite there yet.
As for my credentials, suffice it so say that I do not need to resort to a nebulous and abstract mass of experts to make my case, it’s very evident that there’s a lot of consumer backlash to AI (not necessarily because they hate the tech itself but because it has been shoved into everything without rhyme or reason) and that at the moment the monetary value of it is questionable at the scale it has been invested in. The best consumer use case is image generation, which is very expensive and I bet most people are not willing to pay a sub for it. In fact I do not need to bet, the evidence is out there already as hardly any consumers pay forChatGPT et al, the money they make is overwhelmingly coming from businesses.
Last edited by a moderator:
I'd say iOS 18 Apple Intelligence was more buggy than slop, iOS 26 seems 10x better with its on device models.
However bad Apple Intelligence is I am so grateful they let you shut the entire thing off with a single switch and never nag you about it again if you do turn it off. It's something I wish Microsoft would take the hint for in Windows.
They still have $130 billion in cash and liquid assets that they can deploy at any time. Their competitors are slugging it out losing money on AI. Their AI data centers are being built out using Apple technology. They have home-grown fit-to-purpose AI for specific user tasks. They have home-grown LLMs that are deployed on device. What they are missing is general purpose massive LLMs that most consumers use as glorified search engines.
Like Motorola controlled the Mac destiny, until Apple replaced them with IBM, who controlled their destiny until Apple replaced them with Intel, who controlled their destiny until Apple decided they had a better handle on the future and built their own SoCs?
Like Qualcomm has controlled Apple's destiny until... oops, the C1 is here now.
Like every Sherlocked app has controlled Apple's destiny until it didn't.
Apple has been fairly smart (and lucky) in the way it has used external suppliers while waiting for markets to determine winning and losing technologies. Meanwhile, they've used their r&d budgets to work on things where they can make a case for innovation within a defined consumer market.
Apple's conservative investments in AI are justifiable based on current market dynamics. Measurable profit from AI lies currently in enterprise applications, replacing human agents, and domain-specific worker productivity. These are areas where Apple is not structured to compete. It is instead focused on consumer hardware and services. It does not offer a unified enterprise software platform like those from Microsoft or Salesforce, and it lacks dedicated enterprise-software sales channels. While AI vendors are marketing subscription products aimed at professional users, there is no well-defined consumer AI subscription market. There are some promising ideas for integrating AI into the Apple ecosystem (for example, enhancing Fitness+), but such applications do not require an OpenAI-level commitment to large-scale infrastructure or AGI research.
There’s plenty of demand, just not in the parts of the market where Apple actually plays.
I don’t disagree about AGI. It’s just that Apple needs AGI to make sense for its ecosystem and its customers.
Decades in the computer industry at levels from IC to CEO. PhD in Computer Science from a top school. Seven years in AI research. MBA from a top school.
Last edited:
The difference is that with Apple’s implementation, artificial intelligence will be a feature, not a product. It will be a back-end thing used relatively invisibly to help you do things you want to do without thinking about AI.
Apple’s stumble in this area was last year when they uncharacteristically felt the need to respond to all the hype by rebranding what they were already doing with machine learning and prematurely announcing “Apple Intelligence.” Apple doesn’t usually get caught up with FOMO, but they did with this.
Fortunately, they reined that back in and ultimately didn’t join the crowd by releasing a half-baked hot mess just so they could prance around telling people about their huge AI.
The reality is that the LLM AI being touted by the others (who are supposedly so far ahead) is a monumental waste of energy, hardware and money being used to implement an unsophisticated brute force method for mimicking human language in a deceptive sleight of hand that makes True Believers and the general public think there’s intelligent thought where there is absolutely none.
So Apple will roll out its next-generation machine learning baked into Siri and other Apple products and services in some understated but useful way that functions privately and securely on user devices.
Meanwhile, other companies are going to be left holding their massive investments in huge data centers careening rapidly toward hardware obsolescence as the bubble bursts and the public loses interest in “AI” as a thing, because it will fail to live up to the hype, and as generative AI becomes ever worse as the feedback loop of its own slop makes its output so unreliable that people quit using it.
Being “behind” isn’t so bad a thing when those who are “ahead” are going in the wrong direction.
Apple’s stumble in this area was last year when they uncharacteristically felt the need to respond to all the hype by rebranding what they were already doing with machine learning and prematurely announcing “Apple Intelligence.” Apple doesn’t usually get caught up with FOMO, but they did with this.
Fortunately, they reined that back in and ultimately didn’t join the crowd by releasing a half-baked hot mess just so they could prance around telling people about their huge AI.
The reality is that the LLM AI being touted by the others (who are supposedly so far ahead) is a monumental waste of energy, hardware and money being used to implement an unsophisticated brute force method for mimicking human language in a deceptive sleight of hand that makes True Believers and the general public think there’s intelligent thought where there is absolutely none.
So Apple will roll out its next-generation machine learning baked into Siri and other Apple products and services in some understated but useful way that functions privately and securely on user devices.
Meanwhile, other companies are going to be left holding their massive investments in huge data centers careening rapidly toward hardware obsolescence as the bubble bursts and the public loses interest in “AI” as a thing, because it will fail to live up to the hype, and as generative AI becomes ever worse as the feedback loop of its own slop makes its output so unreliable that people quit using it.
Being “behind” isn’t so bad a thing when those who are “ahead” are going in the wrong direction.
In other news, cold fusion is right around the corner.
Apple's restrained artificial intelligence strategy may pay off in 2026 amid the arrival of a revamped Siri and concerns around the AI market "bubble" bursting, The Information argues.
The speculative report notes that Apple has taken a restrained approach with AI innovations compared with peers such as OpenAI, Google, and Meta, which are investing hundreds of billions of dollars in data centers, chips, and large language model training. This has fueled criticism that Apple is falling behind in the AI space, particularly as Siri has significantly lagged behind more advanced, capable, and reliable conversational systems.
The report argues that market sentiment toward AI spending is beginning to show signs of skepticism, with questions emerging over whether such large investments can be justified by near-term revenue. Against that backdrop, Apple's decision to limit AI-specific capital expenditures has left it with more than $130 billion in cash and marketable securities, giving the company the option to pursue acquisitions or partnerships if valuations of AI startups fall.
Apple's biggest AI-related move in 2026 will be the long-anticipated overhaul of Siri, which is expected to arrive in the spring. The updated assistant is set to be more conversational and capable of completing multi-step tasks. To power it, Apple is believed to be adopting Google's Gemini, reflecting an internal view that large language models may become commoditized and not worth the cost of large-scale proprietary development.
The iPhone is said to be a key strategic advantage. Unlike AI companies that rely on standalone apps or web services, Apple can distribute AI features directly through software updates and system-level integrations across its devices. Efforts by AI companies to build competing hardware face major challenges in manufacturing, distribution, and ecosystem development, areas where Apple has very strong footholds.
The Information also points to recent leadership changes as part of Apple's effort to refocus its AI work. Siri has been placed under Mike Rockwell, who was responsible for launching the Vision Pro headset, following significant delays to the assistant's overhaul. In addition, Apple's AI chief John Giannandrea announced his retirement earlier in December, with parts of his organization redistributed into product-focused teams amid internal concerns about a lack of clear product direction.
While Apple has a history of early but uneven AI efforts, including the original launch of Siri in 2011, The Information argues that these shortcomings have not materially harmed its core businesses. 2026 may be an inflection point in which Apple's cautious strategy could appear prescient if enthusiasm for large-scale AI spending continues to cool and the company finally delivers a more capable version of Siri.
Article Link: Report: Apple's AI Strategy Could Finally Pay Off in 2026
they don't need magical quality. there are like 10 commands which i tell siri like `turn on living room lights`, if siri would automatically learn to get them properly, it would have covered 95% of my cases and i'd still be pretty happy.
Siri fails executing the right command 70% of the time and drives me crazy. it worked better when it was first released.
Siri fails executing the right command 70% of the time and drives me crazy. it worked better when it was first released.
Have you seen how Gemini can actually integrate with Android? It is light years in front of anything Apple has got.
Try reading again. I did Not say that all AI doesn’t do anything useful. There’s plenty of things that get lumped under the name AI that do quite useful things.
I said that specifically the LLM generative AI models primarily being discussed here don’t have a clear justifiable use as far as I have seen. Aside from technically impressive but rather awkward party tricks, what can ChatGPT, Gemini, etc. actually do that justifies the necessary costs?
For instance, some folks on my design team tried using Gemini to generate images of designs I requested them to work on. The results were useless nonsense. Very pretty looking and impressively quickly made for how pretty looking they were, but useless nonsense all the same and still took longer to create than just doing a good enough but better job without the AI would have.
So, my question stands - can anyone point to something that generative AI is actually doing that is worth the cost?
That’s about what I use Siri for too and about all I need it to do.
And for me it works just fine almost all the time. Rarely fails, and when it does, it’s almost always because of a network connection failure in my house. I’ve seen pretty much the same experience with friends & family who use Siri.
I’ve never personally seen anything like a “%70” failure rate or even in the same ballpark.
Spot on. The removal of Launchpad on Tahoe made me restore my machines back to Sequoia. What were they thinking? Had macOS always worked like this, or even if it harkened back to the Launchpad in some meaningful way (customization), then fine...but all they did here is steal the app launcher from Samsung DeX and force it on the desktop user workflow.
As an existential crises looms on the horizon… A-Inc’s stance is “look how much money we’ve saved by not investing in A.I.
You may be right.
Rumor has it Siri will be fully functional with amazing capabilities when it’s released in the first quarter of 2037.
Enjoy the wait.
Completely agree. I’ve long been thinking that Apple’s AI approach is far superior to the rest. LLMs will be a commodity. Why should Apple waste tens of billions of dollars on something that isn’t even profitable and currently has no chance to be? The AI companies are all losing money because training is expensive but inference is even worse. The more they grow, the more they lose, unlike with regular software. The hyperscalars aren’t making money either but are subsidizing the AI companies. NVIDIA is making money but it all depends on the AI companies buying more chips. Once investors realize that most companies have no path to profitability the whole thing goes belly-up and everyone falls together. NVIDIA and the hyperscalars won’t go bankrupt or anything but their stocks will fall, they will have stranded assets and they will take steep losses on their investments. Meanwhile Apple will be flush with cash, ready to lock-in years long deals on memory and TSMC capacity for dirt cheap prices because there will be a massive glut and the manufacturers will be desperate to have a customer. Apple can also buy up some of the stranded assets (for reduced costs) if they wish, to train their own AIs. And in the meantime Apple can simply license the best LLM, letting other companies shoulder the insane costs. Contrast with Meta, who is running the “metaverse” spending approach again- and will take billions in write-downs at the end, again.
WHEN it pops, the only just outcome would be these guys get thrown in jail for the obvious fraud and stock pumping they’re all doing with their ******** deals. Merely being fired just sets off the next round of the bubble cycle as there are plenty of sociopaths in Silicon Valley waiting for the next grift to hop on.
I think Apple’s approach of “AI” being a *feature* and not a product is likely to bring actual usefulness to people. Meaning it won’t be promised as the next god or the ability to do literally anything (as the fraudsters noted above are literally selling to people in interviews), it will bring refinement to platforms *where it makes sense or an actual difference*.
That said, personally I don’t use anything other than the behind the scenes processes already incorporated into things like Photos. The ability to extract text from any photo I take has been a HUGE quality of life refinement for me personally 🤷♂️
Sure Apple can do inference on their devices but they can’t do training and that is where you need massive data centres.
I'm having trouble visualizing a function consisting of all minima, I suppose a monotonically decreasing function would have no maxima, minima I'm not too sure about. I need an example
all the doom and peeps are going t be silent when it comes out because knowing how Apple has done things in the past - they don't release things until they are ready.