You will need an Apple News+ or Wall Street Journal subscription to read this article.
After the hype has settled in a bit I’m much more upbeat about the near future.
-
I am comfortable with saying aloud that the majority of the rank and file tech users don’t care about who’s in the lead for AI. In fact they are already fatigued by the bleating from the media who can’t stop talking and writing. I think people understand that the media’s job is to attract eyeballs and our job is to guild ourselves and families forward.
-
The UI advancements across all platforms is very welcome. People have been asking for improved customisation for years. The ability to change the UI to suit their aesthetic. To customise their screens to a point and that day is nigh and rather than acknowledge this feat this is ignored in favor of the Tech Press “What about your AI though?” It’s probably just me but I’ve felt like mobile UI have been too heavy. You pause a video for whatever reason and suddenly there’s a huge overlay letting you know the video has been paused lest you suddenly grow amnesia and forgot the almost universal symbols for Play/Pause/FF/RW.
Apple doesn’t need to be first or even ahead. The path to the riches are for those that simplify the complex. Horrible UI/UX is actually the standard. I don’t care if a tool is powerful if it’s cumbersome to wield.
7 Likes
I tend to agree with you. Apple has a commanding lead in the minds of the general public. And some of the promised new iPad features even seemed to be specifically designed to placate podcasters. In hopes they will quit their constant b. . . complaining? 
Right now I’m hoping the pluses will outnumber the minuses (I’m reserving judgment of the new UI until I see how much I can tone it down). But if Apple fails to start generating additional iPhone sales in the relatively near future, and the stock market decides it is no longer a growth stock . . . . .
1 Like
Since I don’t have a WSJ subscription, perhaps you could share what you found interesting about it, or note some salient quotes from the article, @Bmosbacker ?
The gist of the article is that Apple’s recent AI announcements were underwhelming. Apple hasn’t delivered an advanced version of Siri, and major improvements may not come until 2026. The article claims that because Apple relies on selling hardware, it struggles to keep pace in an AI-driven world where cloud services dominate.
You can take their assessment for what you consider it’s worth. I have a mixed reaction. I think there is truth to what is being said, but Apple is an innovative and very wealthy company. I believe they are dedicated to eventually getting this right. It may be slower than we or they desired, but I’m hopeful and cautiously optimistic that they will get there.
I think Apple made a mistake not being a private alternative to the Cloud Vendors.
Do you mean having their own cloud-based LLM?
“Many of the AI features announced were more incremental in our view , and already available through competitor applications” UBS analyst David Vogt
Gemini, Co-Pilot and ChatGPT are not local. I have no idea what he suggest are “features” that term is excessively broad. Text manipulation, image creation, summarization, translation etc these are features but in broad terms more of building blocks. We need the mortar that ties these AI “components” together in a way that is palatable for general consumer consumption.
“That means an AI-powered Siri might not be hitting the market until late 2026, or two years after the rival Google Assistant got its first major overhaul”
Consumer purchases are not currently dictated by AI timelines. The people that want to leverage AI know what it can do currently and few are willing to pony up the monthly fees. They also understand the push to recoup investor funds is not their responsibility. Pushing Beta quality AI upon people for a fee isn’t going to happen. The article references Apple’s tardiness by a comparison to Google’s update that occurred two years ago but silent on estimated revenue/profit from Google being early. Who cares if they arrived two years early to the money pit?
“Apple is running well behind peers in AI in other ways too, due in large part to a difference in business models. Microsoft, Alphabet’s Google, Amazon and Meta Platforms can distribute AI through their massive global networks that were already designed to deliver their core business services. Apple has to distribute AI through its devices, which still account for three-quarters of its annual revenue”
What??? AI isn’t a matter of distribution more so compute power and even then I’d balk at the claim that Meta which has but a single window (Facebook) has more distribution power than Apple which has Millions of Mac and iOS devices that people depend on daily. Amazon and Microsoft have AWS and Azure respectively and I admittedly do not understand how they can be leveraged via AI to good effect for the end user.
“The Problem there is twofold. One is that so-called on-device AI hasn’t yet proved to be a major selling point for products such as PCs and smartphones. The other is that Apple’s lack of it’s own cloud-based AI capabilities leaves the 3 trillion dollar company still in need of powerful allies”
Apple published an entire paper on Private Cloud Compute. Again AI is not a Zero Sum game. OpenAI doesn’t have to lose because Apple, Google or Microsoft finds success.
3 Likes
I think the key use cases for LLMs have yet to be demonstrated, i.e. those in daily use by the majority of the people with any device capable of accessing LLMs.
I fail to see how Siri not leading on this is of any great significance without those use cases.
4 Likes
Voice chatting with ChatGPT blows Siri out of the water. I fired up the free version of Chat GPT on my phone and asked it to provide me with a brief timeline of the major events in photographer Diane Arbus’ life and career. It both read the timeline out to me in a voice that was nearly indistinguishable from a real human’s, and provided me with an exportable text version too.
I then asked for a brief explanation of Mitosis, which it provided, again in voice and text. (It outlined a five stage process beginning with Prophase and ending with Cytokinesis, and pointed out that the latter was not technically part of mitosis.)
As an homage to John Siracusa and his ATP toaster oven reviews for Cards Against Humanity, I then asked for the five top toaster oven brands in the US. It listed the brands, highlighted what feature of each brand gets special mention in reviews, and explained how it put together the list.
In every case I was able to access a list of sources.
Siri’s response to questions one and two was a short sentence or two pulled directly from the Smithsonian Institution and Wikipedia, respectively. The response to the third was a list of websites reviewing toaster ovens. I mean, fine, but it’s not any more convenient than a web search, and not as expansive.
If you’re someone who just wants to ask your phone a question and get an answer (voice or text or both), I suspect you’d prefer ChatGPT to Siri.
Note that within the ChatGPT app, you can choose from four tools: Create an image, Search the web, Run deep research, and Think for longer.
I don’t think LLM’s are the end-all-and-be-all of AI/ML, but they definitely have their use cases, even for casual tech users.
2 Likes
AFAIK Apple is mainly competing against Samsung and Google in the consumer market. And right now almost all of the money being generated by AI is coming from businesses, not consumers. IMO, if their call screening works as advertised Apple will have taken a big step toward leveling up with Google’s on-device AI.
What I’m wondering is how is Apple going to make money from Apple Intelligence? An estimated 90% of existing iPhone users can’t run it. And Apple, who spends around $500,000,000 a year to lease iCloud storage from Google (and Amazon?), doesn’t seem the type to give away a lot of AI time.
I tried something similar recently, asking for a comprehensive list of foldable, step-through e-bikes currently available retail in the UK (It is surprisingly difficult to find or search for such a list by traditional methods - it’s all about SEO and sponsored ads and reviews)
It was next to useless. Even tweaking the prompts, I couldn’t get the best known bike brand (Raleigh) to appear in any list nor any of the BTwin models (from Decathlon a major retailer). I guess it’s because neither of them are reviewed online or on YouTube very often. Leading models would appear and disappear on different iterations of the list. When in frustration I asked chat-GPT “what about the Ampere Alter?” it told me that this was an excellent folding e-bike available in the UK, but it still wouldn’t add it to the list!
There’s a very long way to go before current LLMs are ready for real-world use beyond the cautious use by people who are already experts in a field and who can use the tool without relying on its output and maybe deeply distrusting its “intelligence”
2 Likes
You are not alone:
Until LLMs are just the preprocessor step to a symbolic processor, and not the entire stack, the best advice is to proceed with caution.
LLMs are very polite though.
I didn’t intend to suggest that LLMs (ChatGPT included) were reliable sources of information, but rather that ChatGPT’s voice chat offers a more satisfying experience for people who just want to talk to their phones and get information than Siri currently does. In this particular case, Siri wouldn’t have been able to surface the requested information either.
You and I may think that LLMs, given their flaws, don’t have real-world use cases because they aren’t reliable stores of information or still fail at what, for a human, would be a straightforward task. (To say nothing about the considerable environmental and ethical issues inherent in training and maintaining LLMs … or the sea of derivative AI slop out there … or the threat of deepfakes.) But people can and do turn to unreliable tools all day everyday to get information. LLMs tell lies with confidence, and so do many human of the human voices that have access to our screens.
And I think LLMs do have real-world use cases—anyone who has uses Grammerly or a similar service might find an LLM to be a ready alternative, for instance. It is fun to ask a generative chat model to make an image for you—just because it’s fun doesn’t mean it isn’t a legitimate use case. I was around for the early days of the internet, and plenty of people then were dismissive of its potential for wide adoption.
Just like they do from iCloud. They will give you a (very) limited number of tokens a day for free, and then sell you a subscription to the amount you really need.
1 Like
That would be my guess. And once Apple starts charging users OpenAI will probably want a cut. As I recall 30% is the going rate. 
3 Likes
Apple can feel free to allow me to replace Fitness and News+ in a new Apple One Premier bundle that delivers a good amount of AI goodness.
1 Like
This article shows how much money is sloshing around in this aquarium.
“Meta has offered seven- to nine-figure compensation packages to dozens of researchers from leading A.I. companies such as OpenAI and Google, with some agreeing to join, according to the people.”
I like AI and I find it useful. But as a business proposition, it is currently is is an arena for high-stakes gamblers. This is a phase that markets are willing to lose billions for years on the hope that their horse will win. I see no reason why Apple should enter this. I think it makes more sense to pluck off low-hanging fruit to marginally improve this or that. Work on features that can take advantage of “small” data on device.
Offer convenient access to the products of Google/Meta/OpenAI/Anthropic. These guys are fighting to the death and would probably offer stuff cheaply to the Apple universe to raise their own profiles.
But entering into a wholesale competition with these guys makes no sense to me. Wait to see how this all plays out.
Apple has actual products to sell. They are not just a market play.
3 Likes
I’ve turned Apple AI off it’s so rubbish. It annoys me. Every time I asked it to make my text friendly it just added, “thanks a bunch” to the front. For nearly every other simple query it wanted to use ChatGPT and asked for my API.
1 Like
I agree—I’m not particularly interested in an Apple-branded version of ChatGPT. I appreciate the vision Apple has outlined; the real question is whether they can deliver.
As for the billions being invested in AI, I’m glad to see it. AI is a transformative technology. As I understand it—and I’m no expert (I can’t even create a decent shortcut! 
)—AGI matches human ability, while superintelligence (SI) exceeds it in nearly every respect. The core issue is the definition of intelligence. I suspect an Orwellian approach will be taken, in which “intelligence” is redefined downward so that success can be declared and profits made. Even so, the pursuit of AGI and SI will likely move the field forward in ways that are helpful and, when used morally, transformative for good.