Apple Intelligence or ChatGPT: What’s Powering Apple’s New AI?
Toggle Dark Mode
Earlier this month, Apple fired a shot that was heard around the world when it showed off “AI for the rest of us” at its Worldwide Developers Conference (WWDC). No longer were AI systems merely intriguing chatbots and abstract concepts; Apple showed us how it would make AI practical and valuable for people’s everyday lives.
Apple’s answer to this came in the form of Apple Intelligence, an AI system that it built from the ground up to serve the needs of iPhone, iPad, and Mac users with a strong focus on “personal context” — ferreting out information from your devices that’s relevant to your everyday needs rather than merely answering questions, creating images, and rewriting blocks of text.
To be clear, Apple Intelligence will do some of these things, too, but the real magic is in its ability to dig into what’s stored on your device, putting potentially complicated pieces of information together to provide accurate answers to simple queries.
For example, as Apple shared during its WWDC keynote, you’ll be able to make a request to Siri like “play the podcast my wife sent me the other day,” and Apple Intelligence will be able to figure out who your wife is, and then check your email and messaging apps to look for a podcast file or link that you recently received from her.
However, Apple Intelligence goes beyond that relatively one-dimensional example, tying together even more information so you can make a straightforward request and leave it up to Siri to work out the details.
Apple Intelligence is grounded in your personal information and context, with the ability to retrieve and analyze the most relevant data from across your apps as well as to reference the content on your screen, like an email or calendar event you’re looking at.Craig Federighi, Apple Senior VP of Software Engineering
During the WWDC keynote, Apple’s Craig Federighi gave the example of opening a work email notifying him of a rescheduled meeting and then asking Siri if he’ll still be able to make it to his daughter’s play on time.
From that simple query, Apple Intelligence figures out who his daughter is, locates and reads the PDF brochure she sent to him several days ago to get the time and location of the play, checks his calendar to see where the rescheduled meeting is being held and when it will end, and then analyzes the traffic between the office and the theatre to see if he’ll be able to get there in time.
All that happens behind the scenes, allowing Siri to return a simple and concise response to let him know if there will be a conflict between the meeting and the play.
Naturally, this level of data analysis requires your iPhone to delve into some pretty personal information, which is why Apple will use on-device processing as much as possible, only handing off to external servers when necessary. However, this made some folks nervous when Apple later mentioned that it was partnering with OpenAI to integrate ChatGPT into its new Apple Intelligence system.
Apple Intelligence vs. ChatGPT
During the WWDC keynote, Apple made it clear that ChatGPT was an addition to Apple Intelligence. Of course, we all lead busy lives, and not everyone has time to watch Apple’s keynotes, so it’s unsurprising that some folks don’t understand the line between Apple Intelligence and ChatGPT and believe that OpenAI will handle everything.
Part of this may also stem from a misconception that Apple was behind the curve in generative AI technology and would need to rely more heavily on OpenAI and its ChatGPT-4o large language model (LLM). However, Apple didn’t even mention ChatGPT until the end, stressing that it was an extra layer of integration to handle “world knowledge” queries.
Nothing else Apple showed in the keynote had anything to do with OpenAI.
Apple’s keynote lasted nearly one hour and 44 minutes. The first hour covered all the non-AI features coming to iOS 18, iPadOS 18, macOS Sequoia, watchOS 11, and more. Apple took the last 35 minutes of the keynote to unveil Apple Intelligence; ChatGPT only came up for two minutes near the end. During that short segment, Craig Federighi also added that ChatGPT is just the start; Apple plans to partner with other AI chatbots to provide customers with even more options.
The other 33 minutes were about Apple Intelligence features powered entirely by Apple’s own LLMs, with on-device processing mixed with Private Cloud Compute servers run solely by Apple and designed for extreme privacy. This includes the aforementioned personal context features, Siri improvements, writing tools like summarizing and proofreading text and image generation features like Genmoji and Image Playground. That stuff is all 100% Apple.
In one fell swoop, Apple proved it wasn’t nearly as far behind in AI as most pundits thought. In typical Apple fashion, it’s been quietly working on the technology behind the scenes for years but wasn’t about to release something until it was big enough to blow everyone away. We saw that with Apple silicon in 2020 and the Vision Pro last year. Even the 2007 iPhone was an example of that. It seems that Apple Intelligence is no different.
Of course, as the saying goes, the proof of this will be in the pudding. So far, we’ve seen some wonderful demos of what Apple Intelligence should be capable of, but staged demos are just that. When Steve Jobs showed off the first iPhone, it was incredibly impressive, but we only found out years later that it was a carefully scripted demo of a prototype device. Apple delivered on its promises with the iPhone, and hopefully, it will do the same with Apple Intelligence, but it’s not even available in the iOS 18 betas yet; we’ll have to wait and see what it can do once it’s in our hands.
It also remains to be seen how many requests will need to go to the cloud, although those that do won’t be logged or even saved. Apple has promised that third-party security firms will audit its entire Private Cloud Compute infrastructure to guarantee this, and early reports indicate that they won’t even have persistent storage, making it impossible for any data to be retained once it’s been processed.
However, the Private Cloud Compute servers are also merely a fallback for things that go beyond the capabilities of Apple silicon. Apple intends for the majority of Apple Intelligence processing to be done entirely on-device — a strategy it’s employed since it added facial recognition to the Photos app in iOS 10. That’s why you’ll need an iPhone 15 Pro or later (or an M-series Mac or iPad) to handle it.
ChatGPT is in the mix simply because Apple recognizes that its LLMs have one significant blind spot. They’re designed to know everything they can about you (based on what’s stored on your devices), but they’re not yet fully trained to know everything about the world around you.
Still, there are other artificial intelligence tools available that can be useful for tasks that draw on broad world knowledge or offer specialized domain expertise. We want you to be able to use these external models without having to jump between different tools, so we’re integrating them right into your experiences — and we’re starting out with the best of these, the pioneer and market leader, ChatGPT from OpenAI, powered by GPT-4o.Craig Federighi
One of the key places where ChatGPT will be tied into the system is Siri, letting it hand off requests it can’t answer on its own to ChatGPT. Examples include things like asking for recipe suggestions based on ingredients you have at hand or pulling up a photo and asking for decor recommendations.
Apple wants to make sure it’s abundantly clear when ChatGPT is being called upon, which means users will always be prompted to explicitly allow ChatGPT. There’s no possibility of accidentally interacting with it, and from everything we’ve heard, Apple doesn’t intend to provide a way to bypass this prompt — you’ll have to grant permission every single time. That might even drive some ChatGPT fans to just keep using the standalone app.
However, Apple’s ChatGPT integration comes with an additional privacy benefit in that its arrangement means OpenAI will never get any personal information about you — not even your IP address — unless you’re signed into a ChatGPT account (which isn’t required). Apple has built an anonymization layer similar to the concept of iCloud Private Relay, and OpenAI has promised Apple it won’t store any user data for anonymous users. ChatGPT fans and paying subscribers will be able to integrate their existing ChatGPT experience, but those who prefer to remain anonymous will be able to use ChatGPT without sacrificing any privacy.
