Craig Federighi talks about the OpenAI partnership in a new interview

One of the major new features announced with iOS 18 and macOS 15 at WWDC 2024 is Apple Intelligence – a set of AI-based tools. Although these features won’t be available to users until later this year, Craig Federighi, Apple’s head of software engineering, discussed the future of Apple Intelligence in an interview with Fast company.

Craig Federighi talks about Apple’s partnership with OpenAI

Using Apple Intelligence, users can now ask the system to paraphrase text, summarize messages or emails, and create images and even emojis. Siri has also been updated with AI, now understanding the context of what’s on screen and allowing users to control more aspects of the device.

Apple uses on-device processing and cloud processing for Apple Intelligence features. The company has also partnered to integrate ChatGPT into Siri. Although Apple has put a lot of effort into creating its own language models, Federighi acknowledges that there are other good LLM programs available out there.

“These very large parametric models have interesting capabilities that some users appreciate, and we have seen that integration in our experiments can [those capabilities] Much more accessible than [they are] The executive said today. According to Federighi, GPT-4o is one of the best LLM software currently available, which is why Apple wanted to integrate it into Siri.

To make things transparent to users, Siri will ask for permission to send requests to ChatGPT every time Apple Intelligence can’t answer a question. Federighi also says that Apple plans to add support for more third-party language models in the future, so users can choose what works best for them.

Apple Intelligence in China

China is one of Apple’s most important markets, and also one of the most regulated. These regulations make it very difficult for non-Chinese companies to introduce their AI models there – but how does this affect Apple? Without providing many details, Federighi confirmed that the company is trying to “find a way” to bring Apple Intelligence to China.

“We don’t have a timing to announce it right now, but it’s definitely something we want to do.” Apple Intelligence will first be available in US English, but it’s unclear if users in other regions will be able to try it out.

More from Craig Federighi’s interview

A portion of the requests sent to Apple Intelligence are processed by Apple’s own AI servers called Private Cloud Compute, or PCC. The company has found a way to store data requests and destroy them encrypted after they are processed, and Apple doesn’t see anything.

Although Federighi hopes future chips will be able to run larger language models, he points out that having an online model is still important for providing up-to-date information.

“I can’t rule it out,” he says, “but even in that world, I think you would expect that sometimes your device, in serving your request, would reach out to knowledge stores that exist at least. outside of the device. So, even in that future, I think it would be There is a role for contacting external services.

Be sure to check out the full interview on Fast company website.