by Jason Snell
Gurman details Apple two-phase AI rollout
Bloomberg’s Mark Gurman is back from some time off with a blockbuster report about how Apple’s planning on rolling out its new Google-based AI models and functionality:
The previously promised, non-chatbot update to Siri — retaining the current interface — is planned for iOS 26.4, due in the coming months. The idea behind that upgrade is to add features unveiled in 2024, including the ability to analyze on-screen content and tap into personal data. It also will be better at searching the web.
In other words, Apple’s first plan is to make good on all of its broken AI promises from WWDC 2024, using a currently-available Google AI model. It’s an interesting decision, and suggests that Apple’s executives feel those promises hanging over their heads even now.
Gurman continues:
The chatbot capabilities will come later in the year, according to the people, who asked not to be identified because the plans are private. The company aims to unveil that technology in June at its Worldwide Developers Conference and release it in September.
The report says that iOS 27 will feature a newer Google-based model, and it will power Siri in both a voice and text-based chatbot mode. This just makes sense. But Gurman reports that Apple hasn’t committed to launching a full-fledged Siri app in the style of the Gemini, Claude, and ChatGPT apps. This suggests that Apple may be reluctant to embrace the free-form prompt approach, which has its pros and cons. (I think having a place to refer to past chats and continue them is interesting; I also think that making a blank text box the primary interface for anything is a step backward, not forward.)
One last tidbit from Gurman:
In a potential policy shift for Apple, the two partners are discussing hosting the chatbot directly on Google servers running powerful chips known as TPUs, or tensor processing units. The more immediate Siri update, in contrast, will operate on Apple’s own Private Cloud Compute servers, which rely on high-end Mac chips for processing.
I wonder what technical roadblocks are bringing an issue like this to the forefront. Can Apple’s carefully architected Private Cloud Compute infrastructure not provide enough power to run the Google-designed models they need? Will Google host that chatbot on hardware with similar privacy protections, or would this be a crack in Apple’s privacy approach? It will be interesting to see what Apple will do if forced to choose between privacy and functionality.