Apple’s complicated plan to improve its AI while protecting privacy
1 min read
Summary
Apple has reportedly found a way of improving its AI models without needing to train them using its customers’ data, a blog post by the company claims.
Instead of using real data, the plan is for Apple devices to compare a synthetic dataset to samples of emails and messages from opted-in users to its Device Analytics program.
The device determines which synthetic inputs are closest to the real sample, which it signals to Apple to improve its AI text outputs such as email summaries.
The system protects user privacy as the data doesn’t leave the device and Apple can’t pinpoint specific users’ data, merely using the most popular fake samples to improve its AI.
The new system is being introduced in beta versions of iOS and iPadOS 18.5 and macOS 15.5.