Apple gets full Gemini access to build on-device AI models for smarter Siri

New details on Apple-Google AI deal revealed, including Gemini changes: report - 9to5Mac

Today in The Information's latest 'AI Agenda' column, a variety of details of Apple and Google's AI partnership have been outlined. The report says that Apple's partnership is "deeper than previously known." As part of the agreement, Apple reportedly gets "a lot more freedom with Google's tech" than was expected. From the report: Apple has complete access to the Gemini model in its own data center facilities. Apple can use that access to produce smaller models that power specific tasks or are small enough to run directly on Apple devices so they can run the tasks faster, said a person who has direct knowledge of the arrangement. As explained by The Information, this process of producing new, offshoot models is called "distillation." It's a way in which one LLM can transfer knowledge to smaller, more streamlined models. What's the benefit of this? Smaller models can be more efficient, requiring less computing power and delivering results faster. In other words, they make a lot more sense for running on-device than larger models. Since Apple has full access to Gemini, its student model can also learn to imitate the internal computations that Gemini uses to arrive at its answers, which can be more effective than just imitating the answers it spits out. This results in smaller models that roughly approximate the performance of their state-of-the-art teachers but require significantly less computing power to run. The Information's source said that this can be a tricky process, since Apple's goals for Siri don't always align with Gemini's specialties. They also reiterate what we've heard before about Apple's Foundation Models team not giving up on in-house models. However, it sounds very unclear what the AFM team's current goals are. How are you feeling about Apple's AI deal with Google? Let us know in the comments.

Apple Can Create Smaller On-Device AI Models From Google's Gemini

Apple has full access to Gemini to customize the model for Siri and other AI features, reports The Information. Google gave Apple "complete access" to the Gemini model in its own data centers, and Apple can use the access for distillation, or creating smaller models for specific tasks. Apple is able to design models that are built to run on Apple devices without the need to connect to the internet. The Information explains that Apple can ask the main Gemini model to perform a series of tasks that provide high-quality results, with a rundown of the reasoning process. Apple can feed the answers and reasoning information that it gets from Gemini to train smaller, cheaper models. With this process, the smaller models are able to learn the internal computations used by Gemini, producing efficient models that have Gemini-like performance but require less computing power. Apple is also able to edit Gemini as needed to make sure that it responds to queries in a way that Apple wants, but Apple has been running into some issues because Gemini has been tuned for chatbot and coding applications, which doesn't always meet Apple's needs. Apple is relying on Google's Gemini models for the smarter, chatbot version of Siri that's planned for iOS 27, but the Apple Foundation Models team is still working on Apple AI models that are distinct from the Gemini models. Siri will be able to do many of the same things that Gemini and other chatbots are able to do, such as answering questions, summarizing information, scanning and understanding uploaded documents, telling stories, providing emotional support, and completing tasks like booking travel.