Improve artificial intelligence while protecting users’ privacy

By introducing a new way to teach artificial intelligence models, Apple is trying to improve the performance of this technology without direct use of user data. In this approach, users compare user data with real samples and only one signal is sent to the company to determine which synthetic data is closest to actual data.

According to Theverge, Apple has announced that it will no longer need to use real users to improve its artificial intelligence performance. In this new method, users who participate in the Device Analytics program compare artificial data with examples of user messages or emails. The device then only sends a signal of which artificial data is most similar to the actual data, without any data being removed.

This information helps Apple more accurate artificial intelligence outputs such as email summary. These changes are being tested in the beta version of iOS 18.5 and iPados 18.5 along with MacOS 15.5. Apple also uses a technique called defamation so that data cannot be attributed to specific people; The method used in the company’s products since 2016.

Privacy Privacy in Apple’s Artificial Intelligence

In the past, Apple used only artificial data to train its models, which sometimes reduced the quality of the responses. But now with this new strategy, the company is hoping to improve the quality of its artificial intelligence services without making users safe.

Apple’s new approach to teaching artificial intelligence models is a clever effort to maintain balance between privacy and the quality of artificial intelligence outputs. As the AI ​​competition becomes more intense, Apple’s policy can become a strong point against competitors, often relying on users’ data.