Apple says it has found a way to improve its artificial intelligence models without training them on user data or even copying it from their iPhones and Macs. In a blog post first reported by Bloomberg, the company outlined its plans to have devices compare a synthetic dataset with samples of recent emails or messages from users who have connected to the Device Analytics program.
Apple’s devices will be able to determine which synthetic inputs are closest to real samples, which they will report to the company by sending “only a signal indicating which option is closest to the data sample.” This way, according to Apple, it does not gain access to the user’s data, and the data never leaves the device. Apple will then use the most frequently selected fake samples to improve textual AI results, such as email summaries.
According to Bloomberg’s Mark Gurman, Apple currently trains its AI models on synthetic data only, which could potentially lead to less useful answers. Apple has faced difficulties during the launch of its flagship Apple Intelligence features, as it has postponed the launch of some features and replaced the head of the Siri development team.
But now Apple is trying to remedy the situation by introducing its new artificial intelligence training system in the beta version of iOS and iPadOS 18.5 and macOS 15.5, Gurman reports.
Apple has been talking about using a method called “differentiated privacy” to keep user data private since at least 2016 with the launch of iOS 10 and has already used it to improve the AI-powered Genmoji feature. This also applies to the company’s new AI training plans, as Apple claims that introducing randomized information into a larger data set will help prevent data from being linked to a specific person.
