Apple is introducing a new core feature of iOS 18 in the form of Apple Intelligence (“AI” for short, no doubt a clever marketing ploy to appropriate the abbreviation now), which largely works on devices – and of course this requires power.
So much so that only the current iPhone 15 Pro and 15 Pro Max are the only models capable of running AI. So clearly, if you want to, either you already own one of these current generation Pro devices with the A17 Pro chip and its 16-core Neural Engine, or you’ll buy a new iPhone 16 that meets these requirements by fall to the last.
But is that the only reason Apple Intelligence won’t work on the regular iPhone 15 and iPhone 15 Plus?
This is not a sales gimmick
However, Apple does not want to be accused of such a business ploy, but explains in more detail in an interview why more power is actually needed for satisfactory use of the new AI system.
In principle, the AI would also work on older devices, but not in the desired way, which is why Apple is preventing it.
This is explained by the company’s top executives in “The Talk Show” taped at WWDC 2024. Daring Fireball’s John Gruber discussed Apple Intelligence and other WWDC announcements in front of a live audience with the exec of Apple software development Craig Federighi, marketing manager. Greg Joswiak and Head of AI/Machine Learning John Giannandrea.
Of course, it takes more than two hours, so it’s a good thing that Macrumors has highlighted some of the most important passages on Apple Intelligence. Here is what John Gianandrea says:
So when you run these models at runtime (“runtime”), it’s called inference, and inference of large language models is incredibly computationally intensive. So it’s a combination of the bandwidth of the device, the size of the Apple Neural Engine, and the power of the device to run these models fast enough to be useful. Theoretically, you could run these models on a very old device, but that would be so slow that it wouldn’t be useful.
John Giannandrea, Apple
Possible on “old” devices, but very slow
Presenter John Gruber immediately jumps in: “So it’s not about selling new iPhones? Marketing manager Greg Joswiak chimes in: “No, not at all. Otherwise, we would have been smart enough to just use our latest iPads and Macs, right? »
An M1 chip is enough for iPads and Macs, several generations of which now have at least one integrated chip: the iPad Pro of the 2021 model year, the iPad Air came with the M1 in 2022. The iPad Pro is now at M4 level and the iPad Air at M2. All new Macs since November 2020 are equipped with an M chip.
Only the iPhone requires the latest generation Pro or higher as soon as Apple Intelligence becomes available to the general public. That’s at least the current state, and it also seems understandable so you don’t literally fall asleep waiting for Apple Intelligence to process the inputs.
Apple still working to integrate older devices
Apple’s head of software development, Craig Federighi (see his funniest WWDC moments over the years) adds that the company’s first step with any new feature is to figure out how to bring it back to desktops as much as possible. older devices. But when it comes to Apple Intelligence, hardware is the deciding factor: “It’s a pretty extraordinary thing to run models with this power on an iPhone,” he added.
According to Apple, the processing speed of the iPhone 15 Pro with the A17 Pro chip and the 16-core Neural Engine is up to two times faster than the A16 chip in the iPhone 15 and iPhone 15 More. This means that almost 35 trillion operations can be performed per second.
Federighi also hinted that memory is another aspect of the system required by new AI functions. It is therefore probably no coincidence that all devices compatible with Apple Intelligence have at least 8 GB of RAM. Analyst Ming-Chi Kuo saw the real reason last week, with Apple’s statements suggesting a more complex overall situation.
Related reading:
This article was originally published in our sister publication Macwelt and has been translated and adapted from German.
