The battle for Generative AI has engulfed the tech world over the past year, but Apple has remained on the sidelines thus far. The company is looking to change that in late 2024 when it launches the next version of the iPhone alongside iOS 18. The company is widely expected to jump into the AI arms race by adding new AI-based features for the first time, and a new report states these features will run on a mix of data centers and on-device hardware.
Apple has been widely reported as being on the cusp of entering the AI race with the next iPhone, but it was unclear whether its new AI features would be able to run on-device or in the cloud. According to Bloomberg's Mark Gurman, the company will divide and conquer using a mixture of both approaches, as each has its tradeoffs. Applications that run on-device are much faster and more secure than using the cloud or a data center. But they can't access petabytes of data like a data center can, so they can't be used for broad-ranging or demanding queries.
Gurman says Apple will equip its data centers with M2 Ultra SoCs for the most advanced AI-related tasks. At the same time, more simple functions will be handled on-device, which will be the case not only for the iPhone but also for future Macs and iPads. The report provides an example of a simple AI task: summarizing missed text messages or notifications. Complicated tasks could be generating AI images or summarizing long news articles.
Introducing AI features to Mac hardware will give Apple a reason to encourage users to upgrade, as the company will likely say these features will only work on newer hardware. As an example, Bloomberg states Apple will require at least an A18 chip from the iPhone 15 and an M4 chip for the Mac and iPad, which Apple just announced this week. Apple is also expected to announce an all-new AI-powered version of Siri as well, which may require certain hardware to handle AI-related tasks.
Apple is expected to finally reveal some details about its AI plans at the upcoming World Wide Developers Conference (WWDC), which starts on June 10. There have been reports of Apple developing its own on-device large language model that outperforms the GPT4 model used by ChatGPT, Google Gemini, and Microsoft Copilot. However, other reports suggest Apple might also use Google's Gemini for its on-device model.
