Apple details how it plans to improve its AI models by privately analyzing user data
In the aftermath of criticism of the overwhelming performance of his AI products, especially in areas such as notification Summaries, Apple on Monday detailed How it tries to improve its AI models by analyzing user data privately using synthetic data.
Use an approach with the name “Differential privacy“The company said that it would first generate synthetic data and then the devices of users (on condition that they have chosen to share device analyzes with Apple) with fragments of the generated synthetic data to compare how accurately the models are, and then improve.
“Synthetic data is made to imitate the format and important properties of user data, but do not contain any realized content -generated content,” the company wrote in a blog post. “To put together a representative set of synthetic e -mails, we start making a large set of synthetic messages on different topics […] We then deduce a display, called an embedding, from every synthetic message that records some of the most important dimensions of the message such as language, subject and length. ”
The company said that these embedding will then be sent to a small number of user devices that have opted for device analyzes, and then compare them with an example of E -mails to tell Apple which embedding is the most accurate.
The company said it uses this approach to improve its Genmoji models, and in the future would use synthetic data for Beeldspeeltuin, image wall, memories and writing aids and visual intelligence. Apple said it also users who occur to share device analyzes with synthetic data to improve e -mail entrepreneurship.
