Details of Apple's deep collaboration with Google in the field of artificial intelligence have recently been revealed. It is reported that in the upcoming iOS 27 system, Apple will continue to adhere to its privacy-first principle and focus on local edge-side AI processing solutions. To enhance the experience without compromising user privacy, Apple is adopting "knowledge distillation" technology, using Google's powerful Gemini large model to train its lightweight edge-side AI model.

This "distillation" approach focuses on transferring knowledge from a large model to a smaller one, allowing AI running on Apple devices to achieve performance close to that of a large model while significantly reducing computing and inference costs. For users, this means faster response times for edge-side AI, as well as ensuring that user data does not need to be uploaded to the cloud by default, perfectly aligning with Apple's long-standing emphasis on privacy and security. To further reduce model size for local operation, Apple is actively seeking out relevant startups, and previously even considered acquiring Liquid AI, a company specializing in this area.

However, given the full version of Gemini, which has trillions of parameters and requires extremely high computing power, Apple's own private cloud infrastructure clearly cannot handle all the demands alone. According to the leak, to reduce the pressure on Apple's servers, iOS 27's Siri will forward some user requests to Google Cloud and use an authorized version of the Gemini model for processing.

To maintain its privacy commitment when handling tasks in the cloud, Apple has recently approved the use of NVIDIA's "confidential computing" technology. This technology can encrypt content while the GPU processes data and models. Although this may slightly slow down cloud-based AI queries, it significantly enhances the protection of sensitive information. If this strategy is ultimately implemented, future Apple devices running cloud-based AI functions will deeply rely on the collaboration between Google Cloud and NVIDIA AI chips.