Apple’s AI has faced its share of criticism, particularly around features like notification summaries. 😊 In response, the tech giant is doubling down on privacy while aiming to boost performance. How? By using synthetic data—artificially generated info that mimics real user data without containing any actual personal content.
The process involves creating synthetic messages, emails, and more, then converting them into embeddings. These embeddings capture key aspects like language and topic. Devices of users who’ve opted into sharing analytics compare these embeddings with real data samples, helping Apple identify which models are most accurate.
This method, known as differential privacy, ensures user data remains private. Apple plans to apply this across several features, including Genmoji, Image Playground, and email summaries. It’s a clever blend of innovation and privacy—exactly what you’d expect from a company that values both. 😊
