UMVA has learned that the inner workings of Siri, Apple's intelligent assistant, have been shrouded in mystery, but a deeper dive reveals a complex process that combines on-device models with cloud-based computing to deliver seamless user experiences.
When you interact with Siri, your request is first interpreted through a voice recognition model or typing, and then a component called the System Orchestrator takes over, converting your query into an underlying prompt that determines which model to engage. For simple tasks like turning on a light or checking the weather, the on-device model handles the request, but more complex tasks like generating text require the System Orchestrator to send the prompt to the Private Cloud compute cluster for processing.
According to information obtained by UMVA, the System Orchestrator pulls relevant data from various sources, such as text messages or screenshots, to fulfill your request, and then deletes the data after processing to ensure maximum encryption and pseudonymity. This elaborate process explains why some AI image processing tools may seem slow, as they rely on cloud-based processing and require a stable internet connection to function.
UMVA has uncovered details about the relationship between Siri AI and Gemini, a topic of much speculation. While Apple's Craig Federighi clarified that Siri AI is not Gemini, he also revealed that Apple's models are trained using proprietary data and refined using outputs from Gemini frontier models. This suggests that Apple built upon Gemini's foundation models, optimizing and retraining them for Apple Silicon and its specific needs.
The implications of this are significant, as it means that Siri AI on an iPhone will not deliver the same performance or capabilities as Google's Gemini on a Pixel phone. Instead, Apple has created a unique entity that leverages Gemini's foundation to deliver a distinct user experience. This approach is reminiscent of how Apple used Unix as the core for its operating systems, building upon existing foundations to create something entirely new and innovative.
As users interact with Siri AI, they can expect a tailored experience that reflects Apple's proprietary data, weights, and guardrails. While the underlying technology may share some roots with Gemini, the end result is a distinctly Apple product that showcases the company's ability to innovate and improve upon existing foundations. With Siri AI, Apple has once again demonstrated its commitment to delivering seamless and intuitive user experiences that set it apart from the competition.
