June 12, 2024 – Apple has unveiled its highly anticipated iOS 18 and macOS 15 systems at WWDC 2024, introducing a game-changing feature called “Apple Intelligence” – a suite of AI-based tools. Scheduled to roll out later this year, Apple’s Software Engineering head, Craig Federighi, shared insights into the future of Apple Intelligence in an interview with Fast Company.
Apple Intelligence promises to empower users with a range of practical AI-assisted functions, such as text rewriting, information summarizing, image generation, and even emoji creation. Additionally, Siri, the voice assistant, will undergo a significant upgrade, becoming more intelligent. It will now be able to comprehend the context of on-screen content and offer users more control over their devices.
Notably, Apple Intelligence operates using a combination of on-device and cloud processing. In a significant move, Apple has announced a partnership with OpenAI, integrating the renowned language model, ChatGPT, into Siri. While Apple has been focusing on developing its own language model, Federighi acknowledged the excellence of other large language models (LLMs) available in the market.
“These cutting-edge, large-scale language models possess exciting capabilities that can enhance user experiences. We believe integrating them into Apple products will provide more convenience for our users,” Federighi stated. He emphasized that GPT-4o, one of the top-performing LLMs, was a key reason for its integration into Siri.
To ensure user privacy, Siri will seek user consent before sending requests to ChatGPT for processing, especially when Apple Intelligence cannot answer a query. Federighi revealed that Apple plans to support more third-party language models in the future, giving users the freedom to choose the most suitable model for their needs.
Considering China as one of its most crucial markets, Federighi indicated that Apple is actively exploring ways to introduce Apple Intelligence to the Chinese market. “Although there’s no specific timeline yet, it’s definitely something we want to pursue,” he said. Initially, Apple Intelligence will only support American English, leaving it unclear whether users from other regions will be able to access this feature.
During the interview, Federighi also touched on other aspects of Apple Intelligence. For instance, some user requests will be processed by Apple’s proprietary “Private Cloud Compute” AI server. Apple has developed a method to encrypt and destroy data after processing requests, ensuring user privacy is not compromised.
While Federighi envisions future chips capable of running more extensive language models, he emphasizes the importance of having an online model. This is because it provides users with the latest information. “I’m not ruling out that possibility,” he said, “but even in that future, I think there will still be a role for devices to interact with external knowledge bases when processing user requests. So, even in a fully offline processing scenario, interacting with external services will remain crucial.”
