Focused on User Experience and Privacy

Among the biggest questions surrounding models like ChatGPT, Gemini and Midjourney since launch is what role (if any) they’ll play in our daily lives. Apple is striving to answer this question with its own take on the category, Apple Intelligence, unveiled at WWDC 2024.

Unlike the “bigger is better” approach of many competitors, Apple Intelligence is a more tailored solution designed specifically for Apple’s operating systems, prioritizing a seamless user experience. Apple aims for generative AI aspects to blend seamlessly into the operating system, ideally to the point where users are unaware of the underlying technology.

Keeping the Models Small

The key to this approach is creating smaller models trained on customized datasets for specific functionalities within Apple’s operating systems. This approach is believed to increase transparency around the system’s decision-making process. While Apple Intelligence is more focused than larger models, it can handle a variety of requests thanks to the inclusion of specialized “adapters” for different tasks and styles.

ChatGPT, Gemini, and Third-Party Models

Apple Intelligence is designed to work with third-party models like OpenAI’s ChatGPT and Google Gemini, recognizing that its own models have a limited scope. The system will prompt users to share information externally if it determines a third-party application is better suited to respond. Users can opt out of using third-party platforms or specific features of Apple Intelligence, ensuring greater control over their data.

Private Cloud Compute

Apple maintains the same privacy standards for its servers as it does for its devices. While the system won’t explicitly state whether a query is processed on-device or via a remote server with Private Cloud Compute, users can infer this by disconnecting from the internet. Certain operations, like Image Playground, will always be performed on-device.

Responsible AI and Data Collection

Apple’s models are trained on licensed datasets and publicly accessible information crawled by AppleBot. Websites can opt out of having their content used for training through prompts in their code. Apple emphasizes its commitment to empowering users, representing them authentically, designing with care, and protecting privacy.

Apple’s bespoke approach to foundational models allows for tailoring the system to the user experience while upholding privacy standards. The challenge lies in balancing a frictionless experience with transparency and user control, especially for those concerned about privacy and data usage.

Check out the other AI news and technology events right here in AIfuturize!