Apple enhances AI models with private data analysis

Facing criticism over the performance of its AI products—particularly in areas like notification summaries—Apple has detailed its approach to improving its models through private user data analysis aided by synthetic data.

The tech giant outlined its strategy on Monday, emphasizing its commitment to privacy-focused AI advancements. Using a technique called “differential privacy,” Apple explained that it would generate synthetic data to poll users’ devices—only those who have opted in to share device analytics—to test and refine its models.

“Synthetic data are created to mimic the format and important properties of user data, but do not contain any actual user-generated content,” Apple stated in a blog post.

Apple Uses Synthetic Data to Enhance AI Accuracy

To ensure accurate representation, Apple creates synthetic emails covering various topics before converting them into embeddings—representations capturing essential attributes such as language, topic, and length. A sample of these embeddings is sent to select user devices, which compare them against real user-generated emails to evaluate effectiveness and provide feedback.

Apple confirmed that this approach is currently enhancing its Genmoji models, with plans to expand its use to Image Playground, Image Wand, Memories Creation, Writing Tools, and Visual Intelligence. Additionally, Apple intends to leverage synthetic data for email summary improvements among users who opt into device analytics sharing.

By prioritizing privacy-focused AI development, Apple hopes to strengthen its AI capabilities while maintaining user trust in its handling of personal data.

Image: AppleDsign/Facebook