John Gruber, on Apple Intelligence and the iPhones 16:
These are all good features. But let’s say you never heard of LLMs or ChatGPT. And instead, at WWDC this year, without any overarching “Apple Intelligence” marketing umbrella, Apple had simply announced features like a new cool-looking Siri interface, typing rather than talking to Siri, being able to remove unwanted background objects from photos, a “proofreading” feature for the standard text system that extends and improves the years-old but (IMO) kinda lame grammar-checking feature on MacOS, and brings it to iOS too? Those would seem like totally normal features Apple might add this year. But not tentpole features. These Apple Intelligence features strike me as nothing more than the sort of nice little improvements Apple makes across its OSes every year.
Apple reiterated throughout last week’s “It’s Glowtime” keynote, and now in its advertising for the iPhone 16 lineup, that these are the first iPhones “built for Apple Intelligence from the ground up”. I’m not buying that. These are simply the second generation of iPhone models with enough RAM to run on-device LLMs. LLMs are breakthrough technology. But they’re breakthroughs at the implementation level. The technology is fascinating and important, but so are things like the Swift programming language. I spent the first half of my time testing the iPhone 16 Pro running iOS 18.0 and the second half running 18.1 with Apple Intelligence. A few things got a little nicer. That’s it.
I might be underselling how impossible the Clean Up feature would be without LLMs. I am very likely underselling how valuable the new writing tools might prove to people trying to write in a second language, or who simply aren’t capable of expressing themselves well in their first language. But like I said, they’re all good features. I just don’t see them as combining to form the collective tentpole that Apple is marketing “Apple Intelligence” as. I get it that from Apple’s perspective, engineering-wise, it’s like adding an entire platform to the existing OS. It’s a massive engineering effort and the on-device execution constraints are onerous. But from a user’s perspective, they’re just ... features. When’s the last year Apple has not added cool new features along the scope of these?
Apple’s just riding — and now, through the impressive might of its own advertising and marketing, contributing to — the AI hype wave, and I find that a little eye-roll inducing. It would have been cooler, in an understated breathe-on-your-fingernails-and-polish-them-on-your-shirt kind of way, if Apple had simply added these same new features across their OSes without the marketing emphasis being on the “Apple Intelligence” umbrella. If not for the AI hype wave the industry is currently caught in, this emphasis on which features are part of “Apple Intelligence” would seem as strange as Apple emphasizing, in advertisements, which apps are now built using SwiftUI.
If the iPhone 16 lineup was “built from the ground up” with a purpose in mind, it’s to serve as the best prosumer cameras ever made. Not to create cartoon images of a dog blowing out candles on a birthday cake. The new lineup of iPhones 16 are amazing devices. The non-pro iPhone 16 and 16 Plus arguably offer the best value-per-dollar of any iPhones Apple has ever made. This emphasis on Apple Intelligence distracts from that.
The problem isn’t that Apple is marketing Apple Intelligence a few weeks before it’s actually going to ship. It’s that few of these features are among the coolest or most interesting things about the new iPhone 16 lineup, and none are unique advantages that only Apple has the ability or inclination to offer.7 Every phone on the market will soon be able to generate impersonal saccharine passages of text and uncanny-valley images via LLMs. Only Apple has the talent and passion to create something as innovative and genuinely useful as Camera Control.
I can’t agree more with Gruber’s take. My issue with “AI” (a term which, applied today, seems to refer exclusively to a specific class of generative models) is not the fact that it seems useless or fails to offer a few exciting new ways to interact with our devices, it’s the lack of a “breakthrough effect.”
As someone who has studied and implemented machine learning since long before the 2022 ChatGPT-driven AI hype wave began, my propensity to be unimpressed should be taken into account. That said, I have yet to experience a feature or use case for generative models that has me thinking “This. This is the future.”
Good features, to be sure. I like the idea of smarter Siri. I love the idea of integrating App Intents with AI to create more seamless application interaction and make our own devices smarter. But is any of this outside of the expected progression? Had Apple shipped Jarvis, straight out of Iron Man, then that would be groundbreaking. As it is, they’ve shipped some modest and much overdue updates to their virtual assistant that take advantage of a new paradigm for machine learning.
Gruber makes a great contrast with Camera Control - a feature that I agree is much more Apple and, personally, way more intersecting. I have the 15 Pro and don’t plan to upgrade. I considered it for Camera Control. Even if my phone were not on the Apple Intelligence list, I would not have considered it for those features. Maybe I’ll change my mind after they ship.
Another contrast: Apple Intelligence vs. Apple Vision Pro.
The minute I first used Vision Pro, it became obvious to me that this was something truly special. Computing with floating screens, 3D graphics placed specially, and doing everything with hand gestures and voice commands isn’t “like sci-fi;” it is sci-fi.
Marshaling the hardware, engineering, and software expertise, not to mention the creativity and pure momentum that a multi-trillion dollar company can bring to bear required to create a product like Vision Pro is truly something only Apple could have achieved. Everything else would have seen corners cut or limitations put in place that strike at the very heart of what they sought to accomplish - just look at every other VR/AR platform on the market.1
If I were to bet right now on what concept (forget product lines for a moment) is going to drive the next wave of true sustained growth for Apple, my money would be on spacial computing. The value of the technology for entertainment alone is breathtaking in how untapped it is. Imagine the money to be made selling sideline tickets to the Super Bowl or the potential for full-on movies produced to the level (or beyond that) of Apple Immersive Video.
Meanwhile, AI can write emails for me. I know which gets me more excited - and I know what people will spend more money on.
I specifically say “VR/AR” vs. “AR/VR” or “mixed reality” because every other product seems to be virtual reality first and AR or mixed reality second. I am still stunned by how many flagship VR headsets don’t support color pass-through, let alone anything like the Vision Pro experience.