Apple's approach to AI also emphasizes privacy and user control. Users can opt-out of using third-party platforms and can also control the use of Apple Intelligence on a feature-by-feature basis. The company has also released a whitepaper outlining its principles for responsible AI, which include empowering users, representing users authentically, designing with care, and protecting privacy. However, the company acknowledges that balancing user experience and privacy will be a challenge as it rolls out its new operating systems.
Key takeaways:
- Apple unveiled its own generative AI model, Apple Intelligence, at WWDC 2024, which is designed to seamlessly blend into the operating system and prioritize a frictionless user experience.
- Apple's approach to AI involves creating smaller, more topic-specific models trained on a customized data set, which is expected to increase transparency around why the system makes specific decisions.
- Apple Intelligence can process queries either on device or via a remote server with Private Cloud Compute, with the choice not disclosed to the user as Apple holds its servers to the same privacy standards as its devices.
- Apple's AI models are trained on a combination of licensed data sets and publicly accessible information, with the company releasing a whitepaper highlighting principles governing its AI models, including empowering users, representing users authentically, designing with care, and protecting privacy.