Apple Explains Why Apple Intelligence Exclusivity on iPhone 15 Pro Isn’t a ‘Cash Grab’

We know as much about Apple’s AI, dubbed “Apple Intelligence,” as we’re about to, at least until we get some hands-on time with it (which likely won’t be until next year). Still, you can only experience a better Siri and “semantic” AI if you have an iPhone 15 Pro or iPhone 15 Pro Max. Apple finally offered a (slightly) more thorough explanation of why we’ll never see its upcoming AI features come to older iPhones while promising it’s not just a big scheme to force users to upgrade.

And yet, I’m still wondering why cloud-based AI like ChatGPT won’t get backdated when Samsung already proved it can be done. John Gruber from Daring Fireball (via MacRumors) asked the question we’re all thinking. “Is AI exclusivity on iPhone 15 Pro just a cash grab?” Apple’s AI chief, John Giannandrea, and its main marketing head, Greg Joswiak, said, no, of course not.

Any recent M-series chips found in MacBooks and iPads can access Apple Intelligence. The iPhone 15 Pro’s A17 Bionic chip has a 16-core NPU capable of hitting 35 TOPS (trillions of operations per second). That’s supposed to be twice as much neural capability as the A16 Bionic, restricted to the regular iPhone 15 and iPhone 14 Pro.

According to Giannandrea, it’s due to the “inference”—the runtime of AI–that it is “incredibly computationally expensive” for large language models. According to Apple’s lead developers, a RAM component also determines how well AI runs on-device.

“You could, in theory, run these models on a very old device, but it would be so slow that it would not be useful,” he said.

When Gruber asked if this was a scheme to sell new iPhones, Joswiak denied it, saying, “Otherwise, we would have been smart enough just to do our most recent iPads and Macs, too, wouldn’t we?”

There will be different AI features on the MacBook compared to the iPhone, but the neural capacity of Apple’s ARM-based chips started relatively minimal compared to now. The M1 chip from 2020 had a neural processor capable of 11 TOPS, though it will ostensibly receive some of these new AI features. The M2 could do 15.8. The latest M4 chip on the iPad Pro can do 38 TOPS. How well each of these devices runs the AI is another question, though Apple has reiterated it should be fine.

The A16 Bionic chip’s NPU has a purported 17 TOPS speed, though the base RAM on the iPhone 14 Pro was only 6 GB compared to the 15 Pro’s 8 GB. There are bound to be hardware architecture differences between the two Pro-level phones, but judging by Apple’s comments, the 2 GB of RAM seems enough to restrict AI to the latest and greatest iPhone.

Further, there’s some fuzziness on what exactly “Apple Intelligence” is and what is not. Apple’s VP of software engineering and parkour expert, Craig Federighi, said hyped features like Math Notes or smart script on iPad are not built on any of Apple’s AI models, so they’re coming to older devices.

Earlier this year, Samsung started backfilling some AI features that were once exclusive to the S24 over to the Galaxy S23 line, including the mid-range S23 FE. Features like Circle to Search and Live Translate were backdated to the Galaxy Fold 5 and Flip 5. Note that these weren’t all the new Galaxy AI features on the S24. Yet it’s a sign that some of these features purported to require a lot of neural processing capability could work on older devices, especially since many of these AI models run in the cloud. Samsung has even mentioned how one day it might force users to pay to keep the data centers’ lights on as it processes this energy-hungry AI.

We know Apple’s big AI plans will be a holistic package, impacting everything you can do on your device. Siri is supposed to be the one element linking all this together. Apple wants the AI to access your email and texts and even control your apps. For example, if you ask about your plans with your mother later in the month, it should be able to parse that info through texts and emails.

Apple has promised much of this AI works on the devices and any part that isn’t sent to the cloud. This is supposed to be handled by Apple’s own data centers powered by M-series processors. The Cupertino company is further promising we’ll have the most secure and private cloud computing infrastructure to keep users’ data safe. However, other than the demos shown off at WWDC 2024, we still don’t know which aspects of the AI services will be handled on the most recent Bionic chips and which will require cloud computation.

Or do we? Currently, many AI drafting features are based on OpenAI’s ChatGPT. It will work intrinsically with Apple Intelligence on iOS 18 if nothing else than to get the ball rolling on Apple Intelligence. It’s pretty much the only AI feature we know will be out sooner rather than later, and it will necessarily have to run on the cloud. Apple has previously confirmed with Gizmodo that ChatGPT and all Apple Intelligence will stay on the iPhone 15 Pro.