As Apple prepares Apple Intelligence to jump into Silicon Valley’s AI race, it’s relying on one of its strongest advantages: Its army of 34 million app developers.
IPhone users will get their first taste of Apple Intelligence, the company’s artificial intelligence system, later this month. The company is relying on Apple Intelligence to be the strongest selling point for the iPhone 16, its latest generation of smartphones.
Apple’s AI isn’t as advanced as the state of the art coming out of the most advanced labs, such as rivals like OpenAI’s ChatGPT, Google’s Gemini and Meta’s Llama. Apple isn’t using the biggest models, nor can it pull off some of the more show-stopping tricks of the bleeding-edge voice models — OpenAI’s latest can sing, for example.
Where Apple is hoping to distinguish its AI is that Siri may actually be able to do things on your phone — send emails, decipher calendars and take and edit photos. That’s something other company’s AI chatbots cannot currently do, and to accomplish this, Apple is beckoning its army of third-party developers to fine tune their apps to collaborate with Apple Intelligence. Eventually, Siri may be able to trigger any action in an app that a user can take, part of the company’s long term vision for Siri, Apple said in June.
“Siri will have the ability to take hundreds of new actions in and across apps,” said Apple’s Kelsey Peterson, director of machine learning, in the Apple Intelligence launch video.
Apple can easily make this happen for its own apps, but for Apple Intelligence to interact with the millions of non-Apple apps, it needs developers to embrace a new way of programming their apps. This means developers will need to create as many as hundreds of snippets of additional code called App Intents.
Apple has a strong history of getting its developers to support new platform initiatives, and it’s running a well-worn playbook to get them on board — personal attention from developer relations, a party-like atmosphere at the company’s annual developer’s conference and most importantly, it dangles App Store promotion that can lead to millions of downloads for developers who get on board.
If third-party developers jump on board and the Siri system works as advertised, it could represent one of Apple’s biggest and most durable advantages in the AI race.
“You should be able to string things together and kind of get that future we’ve all been envisioning where you can use Siri conversationally, to do a bunch of things at once,” said Jordan Morgan, an iOS developer who’s written a tutorial about App Intents.
Whether Apple is successful at cajoling its millions of developers is a critical question, and the stakes are high for the company.
The company is relying on Apple Intelligence, which only works on last year’s iPhone 15 Pro or iPhone 16 models that came out this year, to spur a wave of upgrades and boost flat iPhone sales. If Apple’s improved Siri is poorly supported by developers or it fails to impress, it could cool iPhone sales, and customers could wind up choosing to use a rival’s voice assistant through an app instead of the built-in Siri.
Inside the Music app, for example, Apple has built about 10 intents, including actions like “Add to Playlist,” “Play Music,” or “Select Music.” A single app intent should define a single action, programmers say.
If you take a caffeine tracking app, for example, one intent would be the ability to show an overview of exactly how much caffeine the user has logged today, Morgan said.
When that App Intent is finished, Apple’s various “system experiences,” such as widgets, live activities, control center and Shortcuts, will be able to quickly display a current running tracker of how much caffeine has been logged without the user ever opening up the tracking app.
System search is another big draw for some developers. App Intents will allow apps to surface specific emails or other more granular data inside Spotlight, Apple’s system search.
App Intents don’t take that long to write, developers say, often requiring only a few lines of code.
In previous years, Apple recommended that developers adopt App Intents for their most important features, said Michael Tigas, the developer of Focused Work, a productivity app.
“Now, if there’s a way to adjust your app to perform any general action then you should create an App Intent for it,” Tigas said.
Fortunately for developers, they still have time to write all the code necessary for App Intents. While Apple Intelligence is starting to roll out next month, the biggest improvements to Siri aren’t scheduled to be released until next year.
Apple’s new Siri system will better understand questions even if a user makes a speaking error, a direct result of Apple’s work with language models, a relative of the large language models that power systems like OpenAI’s ChatGPT.
That means that Siri will be much more flexible in understanding the hundreds of different ways a user could phrase, for example, “apply a photo filter to an image I took yesterday.”
Apple has to train and test its model to understand the range of the most likely commands and questions for any given category of apps.
A downside to Apple’s approach is that only a few categories of apps will be supported by the new Siri at first, starting with photo and email apps. Eventually, Siri will support apps that focus on books, journaling, whiteboards, managing files, word processing, browsers, camera and photos, the company said.
Developers are already imagining how they might plan for users to interact with their apps with their voices.
A representative for Superhuman, a premium email app, told CNBC that it plans to use Apple’s AI system to enable questions about the contents of emails, such as “Hey Siri, when does my flight depart?” or “Hey Siri, when am I meeting with James to review his proposal?”
There’s a downside to Apple’s plan in the eyes of some developers who worry that users will spend less time inside their apps or confuse Apple Intelligence with the AI features they’ve built themselves.
“If this story were only about App Intents, developers would worry that their products might be reduced to the role of the plumbing that powers Siri, and leave them unclear on how to build sustainable businesses around it,” Igor Zhadanov, CEO under of Readdle, which makes email app Spark, wrote in an email.
Another drawback is that Apple Intelligence features will only be available on the latest iPhones, a small subset of the total iPhone user base. That limited market of iPhone users may discourage developers from investing time and effort into supporting the technology in the near term.
“Apple are limiting these kinds of Apple Intelligence features to the new 2024 iPhones and the expensive models from last year, so you won’t be able to build something for the masses anyway,” Tigas said.