Supply chain analysts have no ability to see code. His opinions on Apple’s AI integrations are meaningless.
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.
Kuo: 'No Sign' of Apple Generative AI Technology Coming in 2024
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple is already deploying AI in its products, but not calling it AI. For example, automatic translation in pictures and magic cropping in pictures. AI, in the ChatGPT/ Midjourny sense of the word, is a flash in the pan that is amazingly hot, and amazingly flashy. "AI" (ok, actually applied statistics) was hyped up as being amazing, and now it's crashing down hard.
Apple will never, ever, release a competitor to ChatGPT or Midjourny. It's too bad for the environment, impossible to develop a "blue ocean", and isn't as good as people believe it is.
I wonder if they can do an autocorrect that's based on machine learning? That should be possible.
Apple will never, ever, release a competitor to ChatGPT or Midjourny. It's too bad for the environment, impossible to develop a "blue ocean", and isn't as good as people believe it is.
I wonder if they can do an autocorrect that's based on machine learning? That should be possible.
And how does that work? In fact, what exactly is "AI" in this context. I consider "AI" to be marketing fluff at this point.INTEL has a separate co processor for AI in Next Years release of the Meteor Lake Processors.
So I dont think ANY of the Mac Chips have a separate AI Co Processor including the upcoming M3 3nm Chips.
So the future with AI and Apple look way way off. Much like Touch Screen Macs.
The Rundown On Intel's Meteor Lake Chip
The details on Intel's 14th-gen processors are largely speculative, but here's what we do know about the upcoming Meteor Lake release.www.howtogeek.com
The funny thing is Apple refers to it as what all this is, Machine Learning.
AI became a stock market buzzword that companies have mainstreamed by slapping it onto products and names to get a temporary pop in their stock price.
What we’re *calling* AI today is nowhere near what we *think* AI is supposed to be, artificial general intelligence. The funnier part is these language models and machine learning are a complete dead end to the pursuit of AGI. AGI will require entirely new techniques and technologies.
False statement that has been proven so countless times.
Apple isn't better than Samsung in AI in the first place anyway.
How is that a false statement? That's how Samsung's "moon mode" works. Someone took a picture of the moon, removed a bunch of information of the picture in photoshop, took a picture of the screen with a Samsung phone camera in "moon mode", and the information reappeared.
source: https://www.theverge.com/2023/3/13/23637401/samsung-fake-moon-photos-ai-galaxy-s21-s23-ultra
It is a false statement.
Here it's explained in detail how the feature works. IT DOESN'T REPLACE THE ORIGINAL PICTURE WITH ANOTHER ONE FROM THE INTERNET.
How Samsung phone camera uses AI for moon photos & pictures | Samsung UK
Find out about the deep learning-based AI processes that Galaxy phone cameras use to take high quality photos of the moon.
www.samsung.com
100% agreed. ChatGPT "AI" (or Applied Statistics) is not just a technology dead end (because of hallucinations), not just a business dead end (because of the inability to create a "blue ocean" and to allow profits to be developed), but is also a dead end because of the ecological environmental harm it generates (source https://www.theguardian.com/technol...l-intelligence-industry-boom-environment-toll).
I think this time next year people would have moved on from this type of AI systems, and attention would be drawn to something else that's new and flashy.
I agree. I want to be secure that my answer is right, and not an hallucination. However, that doesn’t mean that Apple gets a free ride where Siri is concerned. If it takes six weeks to make a change, then hire six more engineers, and every week there should be a change in its capabilities. It doesn’t have to be AI but I think Apple device owners should be able to expect reliable and consistent information from Siri as well as new features. And as I am always saying, tell us about them, don’t expect us to just happen upon them . Apple shold toot Siri’s horn, and not expect someone else to do it for them.
Of course Samsung says it's not replacing information and their system actually works as advertised 🙄. Do you have a better trusted source, or should I just block/ignore you?It is a false statement.
Here it's explained in detail hos the feature works.
How Samsung phone camera uses AI for moon photos & pictures | Samsung UK
Find out about the deep learning-based AI processes that Galaxy phone cameras use to take high quality photos of the moon.
I read your posting history, and to save us both time, I decided just to block you. Bye.
It IS Tim Cook’s biggest failing. If he were not so focused on vision pro, which a large number of Apple users cannot afford, and wouldn’t want if they could, then he might see that this feature, which is available in every device, and is the face of these devices, should be the absolute best thing about all of them.
Of course it works as advertised. I have an S23U and I actually know how it works. I also can take moon shots without Scene Optimized, so no AI, the hardware is perfectly capable and multiple people already proved it.
LOL, official documentation isn't "a trusted source" 🤣🤣🤣. Yeah you can block me alright, you don't have any real arguments anyway.
Great thank you! Sayonara.
I don’t need “generative AI”, I need to be able to communicate with Siri the same way I can talk to ChatGPT.
Is Siri only working well on devices with the neural engine? I have a 2017 iPad as well as an M1 iPad Pro. Siri works comparatively well on the iPad Pro but is absolute garbage on the 2017 iPad.
Used to buy apple products every single year:, Hardcore, now im really considering moving forward giving samsung a chance. The Lack of Apple in certain areas is poor.
The progress of Apple's generative AI technology is significantly behind its competitors and there is no sign that the company plans to launch AI services next year, according to analyst Ming-Chi Kuo.
In a new post on Medium outlining how Apple's imminent earnings report will affect Apple stocks and the supply chain, Kuo explained that the company will likely not dedicate much time to discussion of AI during its earnings call due to its lack of progress in the area. There is reportedly no sign that Apple has plans to launch or integrate AI computing or hardware products in 2024, indicating that AI is unlikely to boost the company's stock price or supply chain in the immediate future.
Last month, Bloomberg's Mark Gurman said that Apple was working on "Apple GPT" artificial intelligence projects that could rival OpenAI's ChatGPT. Apple does not yet have a "clear strategy" for creating a product for consumers, and while it could be planning to make a "significant" AI announcement in 2024, Gurman claims it has no concrete plans as of yet.
During Apple's May earnings call, Tim Cook said there are a "number of issues that need to be sorted" with AI, and that it's important to be "deliberate and thoughtful" in the development approach. Cook also said that Apple views AI as "huge," and plans to "continue weaving it in products on a very thoughtful basis."
Article Link: Kuo: 'No Sign' of Apple Generative AI Technology Coming in 2024
My problems with Siri are not about voice recognition, at least not on my M1 iPad Pro. My problem is Apple giving Siri functions, and then taking them away. Two prime examples are that Siri can no longer search your photos for objects and pets and locations and dates, now it only opens the Photos app now, which is exceedingly not useful when you have thousands of photos. Another example is when it used to be able to add data to a note and they took that function away too. These were very helpful things but I’m sure for those whose speech recognition function is a problem that is indeed more important.
Along with a functional definition of "intelligence".
That's been the problem with AI since the 60's. "Intelligence" remains a mirage always on the horizon because we're driven to believe that humanity holds some special ability that's not easily replicated.
Your problem is your narrow definition of “replace”. If you think Samsung has a series of pngs that it crops and replaces your moon with, that’s not what’s happening. What’s happening is that the same data is encoded in its ”moon fixer” network which makes it easier to shift, reproject, and color match.It is a false statement.
Here it's explained in detail how the feature works. IT DOESN'T REPLACE THE ORIGINAL PICTURE WITH ANOTHER ONE FROM THE INTERNET.
How Samsung phone camera uses AI for moon photos & pictures | Samsung UK
Find out about the deep learning-based AI processes that Galaxy phone cameras use to take high quality photos of the moon.
But if you actually looked at the Verge article that @JustAnExpat linked to, Samsung is creating information that simply doesn’t exist in the source image. That’s not super-resolution or anything of the sort, it’s replacing your picture with data from other, better, pictures.
The staggering energy demands of these systems that we’ve thrown into the public with no clear direction really is a worrisome aspect of all this. It’s takes something like 10x the energy to do an “AI” session than a search does. Really, really concerning given how much they’re trying to drive adoption of it.
I tend to think that’s the main distinction between human intelligence and artificial intelligence at this point: power efficiency. We do a lot with 20W.
Maybe CoPilot in Windows is not that useful. Still, Apple could improve Pages, Numbers and Keynote with generative AI, like MS did with CoPilot in MS Office. What I have seen looks very useful, specially for business and enterprises.
I think the problem is, users that obviously don't own these phones that they hate on(and can't test it for themselves), keep insisting with the same stuff even if months have passed and no new information to reinforced their beliefs, aka "opinions" has appeared. Actually I periodically see random posts that disapprove the hole "moon shots are Fake" drama.
Also he clearly said: that’s not the Samsung type of AI ... means -finding a better one on the internet- and -replacing it- with that. So what narrow definition of “replace” ? He simply said Samsung's AI actually replaces the original photos with 3rd party photos(there's no nuance, that's what he meant) and its obviously not true.
The Verge article is outdated(not to mention users generally weren't able to recreate the same extreme example) and it's premise(what it tries to suggest) was disapprove by multiple user tests.
It's not replacing anything, it recognizes what it sees and enhances it as best as it can(upscale resolution, details), that's why results vary. Google and Samsung have an AI based features that fixes blurred photos, does that mean that the resulting photos are fake? NO, ABSOLUTELY NOT. Samsung has a feature that removes shadows or reflections from photos, does that mean that the result is a fake photo?
Last edited:
Says here Apples neural engine only handles media manipulation. not a true AI Co processor.
What Is Apple's Neural Engine and How Does It Work?
Apple's custom AI coprocessor drives features like media analysis and computational photography on the iPhone, iPad, Mac, and Apple TV.www.makeuseof.com
You are about as clueless as Kuo is on this topic. You posts are flawed on several fronts.
First, Intel's VPU has a media focused AI-augment history.
Intel® Core™ Ultra Processors
The latest Intel® Core™ Ultra processors enable you to use the most AI experiences across desktop, mobile, and edge.
www.intel.com
The 'V' in VPU name comes from 'Vision'. So wagging your finger at the Apple NPU saying it is heavlly tasked with video AI compute tasks is a bit like the pot calling the kettle black.
Second, Apple has both the NPUs and AMX ( Apple Matrix extension in the Arm cores). There are two subsets of matrix works that Apple does well. That holds up very well versus Intel's somewhat hobbled AVX ( no full AVX512 in meteor lake) and VPU ( Intel has a Advanced Matrix Extention also named AMX coming, but won't arrive in clients until after Meteor Lake. I think it is Granite lake first for that. ). Apple's NPU cores are not their only AI silver bullet.
Third, The assertion that intel has a "separate AI chip" is is not really true. The SoC chip in Meteor Lake handles a variety of stuff. The DRAM Memory Controller, security , two "low power cores" (even lower power than the 'E' cores all in addtion to this VPU. Probably not really aimed at most user level apps at all. Mostly background stuff when the system is idle/'sleeping'.) and a few other things ( Display Controller, Media handling (AV1, video processing/inferencing, etc. ).
Some overview here
Hot Chips 34 – Intel’s Meteor Lake Chiplets, Compared to AMD’s
During a presentation at Hot Chips 34, Intel detailed how their upcoming Meteor Lake processors employ chiplets. Like AMD, Intel is seeking to get the modularity and lower costs associated with usi…
chipsandcheese.com
and here
Intel Disaggregates Client Chips with Meteor Lake in 2023
At Hot Chips 34, Intel shared details on its upcoming 2023 Meteor Lake processors that will be composed of disaggregated tiles
So it really is not a whole chip dedicated to AI. Nor is the TOP ( tera-ops-per-second) really going to be all that much better than the NPU cores or better .
[ There are three other chiplets. Compute -- CPU cores ( P and E ) , GPU -- GPU cores I/O -- Thunderbolt controllers , USB , PCI-e , Ethernet/WiFi/etc. , embedded DisplayPort , etc. ]
Intel isn't really 'buying' a whole lot here by using a different die . Apple is not 'loosing' by not going chiplets. Intel is going to pay a bigger Perf/Watt loss penalty than Apple is. Apple is going to make a bigger TSMC chip than Intel's TSMC chips. There probably will not be some massive "AI processor" die area allocation loss here between the two.
Compared to the likely contemporary Apple M3 the Intel SoC will be using an older TSMC process.
"...
| Intel Meteor Lake Tile/Chiplet | Manufacturer / Node |
| CPU Tile | Intel / 'Intel 4' |
| 3D Foveros Base Die | Intel / 22FFL (Intel 16) |
| GPU Tile (tGPU) | TSMC / N5 (5nm) |
| SoC Tile | TSMC / N6 (6nm) |
| IOE Tile | TSMC / N6 (6nm) |
Intel's Meteor Lake, Its First PC Chips With TSMC Tech, Launch This Year
'Intel 4,' TSMC 5nm and 6nm, all living in harmony in one chip.
So unlikely going to bet Apple on core count per unit area allocated. The M2 is on TSMC N5P. and the M3 is likely on some variant of N3. The VPU of Intel's is going N6 ( and Apple has not been on a N7 family process for several YEARS).