Apple Confirms iPhone's AI Limitations
Apple's AI suite, Apple Intelligence, is not available on most existing iPhones. The limitation is due to the computational expense of running large language models and the need for sufficient processing power. Apple's strategy of keeping RAM levels low has historically optimised iOS but limits AI capabilities on older iPhone models.
However, there was a surprising revelation - Apple's AI, known as Apple Intelligence, would not be available on the majority of existing iPhones. This decision was explained during the "Talk Show Live From WWDC" hosted by John Gruber, where Apple's management team shed light on the reasoning behind this limitation.
John Giannandrea, one of Apple's executives, addressed the issue of restricting Apple Intelligence AI software to the iPhone 15 Pro and 15 Pro Max. He explained that running large language models, known as inference, is computationally expensive. Factors such as device bandwidth, the size of the Apple Neural Engine, and the device's processing power all contribute to the need for newer iPhone models. Giannandrea stated that older devices could theoretically run these models, but the performance would be so slow that it would render them useless.
Looking back at the history of the iPhone, it becomes evident that Apple has always aimed to keep RAM levels as low as possible. This strategic decision allowed iOS to have an advantage over Android, as Apple's tight integration of software and hardware enabled optimisation for a single footprint. On the other hand, Android had to cater to various chipsets and I/O devices, necessitating more memory.
While this approach has served Android well in the era of AI smartphones, Apple faces limitations in back-porting its AI suite to older iPhone models. Samsung, for example, has introduced Galaxy AI not only to its latest S24 family but also to older models like the Z Fold 5, Z Flip 5, and Galaxy S23 family. However, even Samsung cannot go back further due to the intensive processing power required for generative AI. Qualcomm, a major Android chipset manufacturer, has added the necessary hardware to its current chipset to support AI capabilities.
Interestingly, Apple will be able to roll out its AI to the 2020 M1-powered MacBook Air and the 2021 iPad Pro, both of which utilise the same architecture as the iPhone's A14 Bionic chip. Both chipsets feature Apple's Neural Processing Unit (NPU) and offer 11 TOPS (trillion operations per second) over 16 neural cores. This indicates that Apple's decision to prioritise AI expansion on its tablet and laptop rather than its smartphone has resulted in a loss of AI capabilities on mobile devices.
Apple now finds itself playing catch-up in the generative AI race. Google kickstarted the smartphone AI race in October 2023, while Microsoft recently introduced AI to Windows 11 through its Copilot+PC program. Additionally, various online services, such as Grammarly, have leveraged AI to enhance their offerings. Given the potential lead Apple could have had in AI if every iPhone from 2020 onwards supported Apple Intelligence, it is likely that Tim Cook and his team would reconsider their decision.
Apple's AI suite, Apple Intelligence, is not available on most existing iPhones.
The limitation is due to the computational expense of running large language models and the need for sufficient processing power.
Apple's strategy of keeping RAM levels low has historically optimised iOS but limits AI capabilities on older iPhone models.
Source: FORBES