Apple announced a new AI training method that utilizes synthetic datasets compared to user data samples, ensuring privacy as no personal data is sent to its servers. Devices will relay only signals indicating which synthetic data is closest to the users’ data, allowing for improved AI outputs like email summaries. The approach aims to overcome challenges the company faced with its AI features and is set to be introduced in upcoming beta versions of its operating systems. Apple emphasizes user privacy, leveraging differential privacy methods to protect individual identities.
Apple’s new AI training system will utilize synthetic datasets compared to user samples without accessing user data, enhancing model accuracy while maintaining privacy.
The company aims to improve AI capabilities by having devices relay signals on synthetic inputs, ensuring user data does not leave devices in the training process.
Collection
[
|
...
]