Facing criticism over lackluster performance in its artificial intelligence features, particularly regarding notification summaries, Apple outlined a new strategy to enhance its AI models by privately analyzing user data with the aid of synthetic data.
Apple plans to utilize a method called “differential privacy” to carry out improvements, aiming to strengthen privacy protections while refining product quality. According to the company’s explanation, the process begins with the generation of synthetic data—artificially produced datasets designed to replicate the structure and key characteristics of actual user information without including real user-generated content.
These synthetic datasets will first undergo algorithmic processing to create specialized representations known as embeddings, which encapsulate critical attributes such as language style, topic relevance, and content length. Once developed, a select portion of these embeddings will be distributed to a limited set of user devices, specifically those whose owners voluntarily agree to share device analytics with Apple.
These user devices will automatically compare the synthetic embeddings against a user-generated sample of data, such as email messages, providing Apple with insight into which embeddings most accurately resemble genuine user information. This feedback will subsequently guide Apple in making enhancement adjustments to its AI algorithms.
Presently, Apple is deploying this improved synthetic-data strategy specifically to boost the capabilities of its Genmoji models. In the future, the company plans to expand utilization of synthetic data analyses to additional products including Image Playground, Image Wand, Memories Creation, Writing Tools, and the Visual Intelligence suite. Additionally, Apple intends to apply these methods in refining its email summary features, again relying only on the input of users who explicitly opt into the analytics program.
