Apple is now adopting a new strategy to train its AI models while upholding its strong privacy commitments in order to enhance the AI-backed features. The company is altering how it uses artificial intelligence (AI) to improve tools like Siri and email summarisation in future software updates, according to a recent blog post by the Apple Machine Learning Research site and a Bloomberg report.
Apple has historically trained its AI models using artificially generated content, or synthetic data. There are certain drawbacks, particularly when training models to perform complex tasks like long-form summarisation, even though this has removed the need to analyse actual user data.
Also read: iPhone 16 Plus price drops by over 14,900 on Flipkart: How this deal works
Apple has created a new privacy-preserving system to address this issue, which will use tiny samples of recent user emails on devices that have chosen to participate in Device Analytics. According to Apple, the procedure doesn’t give it access to user identities or specific emails. Rather, it employs a technique known as embeddings, which represents emails according to their language, topic, and length. These are compared with synthetic messages on the device.
Also read: OnePlus 12 price drops by Rs 19,001 on Amazon, here’s how to grab this deal
Without ever seeing the real emails or knowing which device chose what, Apple is able to determine which kinds of fake messages most closely resemble typical communication patterns thanks to differential privacy. In the end, this will enhance how AI features create or summarise content across all applications, including Mail and Notes, and assist Apple in improving its synthetic training data.
This is expected to be released in the next beta versions of macOS 15.5, as well as iOS 18.5. The precise rollout timeline and other specifics of this “innovation,” however, are still unknown.