At Apple, AI stands for “Apple Intelligence” and is being implemented in everything • The Register

Machine Learning


The world's largest developer conference Speaking from the rooftop of Apple's headquarters at the Worldwide Developers Conference on Monday, CEO Tim Cook declared that the iPhone giant's products now come equipped with “advanced” intelligence capabilities.

Apple has been slow to incorporate what it generously calls “artificial intelligence” into its products, but it would have the world believe it's just being hip and cautious about its “Apple Intelligence” machine learning, which is based on personal data and privacy.

Apple has a history of taking ideas and refining them into platform products: its Macintosh borrowed innovations from Xerox PARC and commercialized the graphic interface, its iPod/iTunes combination transformed the MP3 player market and digital music sales, and its iPhone changed the way people think about mobile devices, computers and the Internet.

Similarly, the company's advertising draws on the past: just as Cupertino's poorly received “Crushed” ad for the iPad Pro was reminiscent of LG's 2008 ads for its KC910 Renoir phone, Apple's WWDC opening featuring skydiving executives recalled Google's rather impressive Project Glass skydiving stunt at Google I/O 2012.

iCompany's approach to artificial intelligence follows a similar storyline: letting early market entrants like Microsoft fail before coming up with a more sophisticated approach.

“Apple Intelligence is a personal intelligence system with powerful generative modeling built at the core of iPhone, iPad and Mac,” said Craig Federighi, SVP of Software Engineering, “to provide the intelligence that's most helpful and relevant to you based on your personal context — while protecting your privacy every step of the way.”

Apple has been focusing on enabling specific AI features across various applications: iOS 18, iPadOS 18, and macOS Sequoia all include a framework called Writing Tools, for example, that allows users to use machine learning to rewrite, proofread, and summarize text in Mail, Notes, Pages, and third-party apps.

The Mail app finally has Gmail-style message categorization, categorizing messages into Primary, Transactional, Update, and Promotional views and highlighting high-priority messages (again using machine learning). High-priority notifications are also flagged this way. And the Notes and Phone apps can record, transcribe, and summarize audio.

Apple Intelligence has also been applied to image generation, where you can now create images on demand through Image Playground, an image generation service that will appear in apps like Messages and Notes, and Photos will have a clean-up tool to remove unwanted elements from photos.

Siri also gets Apple Intelligence built into it, claiming the venerable voice assistant service has better data awareness across apps, allowing for better interactions. As a signal of that expanded reach, when you launch Siri on your iPhone, a visual effect appears that surrounds the entire screen.

In software that Apple plans to release for iOS, iPadOS, and macOS, Siri users will be able to type commands as well as speak them, and Siri should be able to execute commands in either mode while maintaining context. So if you query the weather in a location followed by a command to schedule an event “there,” Siri should infer that “there” means the location mentioned in the previous interaction. Siri will also be able to orchestrate hundreds of actions across multiple apps, such as taking a picture from the Photos app and adding it to an existing entry in the Notes app.

According to Federighi, “The foundation of our personal intelligence system is on-device processing. We've deeply integrated it across your iPhone, iPad, Mac and apps, so it can recognize your personal data without collecting it.”

But performance depends in part on the size of the model, and not all machine learning systems can fit on a mobile device, so Apple has come up with a privacy-focused, cloud-based model.

“We want to extend the privacy and security of iPhone to the cloud, giving users even more intelligence,” said Federighi. “So we developed Private Cloud Computing, which gives Apple Intelligence flexible and scalable computing power, allowing us to access even larger, server-based models for more complex requests, all while protecting user privacy.”

Federighi said that when a user makes a request to use an AI model, Apple Intelligence determines whether the request can be processed on the device, and if not, it processes it privately on servers in Cupertino.

“Your data is never stored by or accessed by Apple,” he says. “Your data is used only to fulfill your requests. And just like your iPhone, independent experts can inspect the code that runs on these servers to verify this privacy promise. In fact, Private Cloud Compute cryptographically ensures that your iPhone, iPad, and Mac refuse to communicate with our servers unless the software is publicly logged for inspection.”

Apple Intelligence is available for free on iOS 18, iPadOS 18, and macOS Sequoia, but Federighi said Apple plans to support other artificial intelligence tools as well, starting with OpenAI's ChatGPT.

You can access ChatGPT for free without creating an account

“You can access ChatGPT for free without creating an account,” Federighi explained. “We don't log any requests or information, and ChatGPT subscribers can connect their accounts to access paid features within our experience.”

Federighi said he checks with users before sharing any information and that the company plans to support other models as well.

Because WWDC is a developer conference, Federighi noted that developers will be able to access Apple Intelligence features through a variety of updated SDKs and new APIs. “For example, developers can add an Image Playground experience to their apps with just a few lines of code,” he said.

For example, the extension of the App Intents framework makes it easier to use related features such as Apple Intelligence Services and Siri to handle automated tasks across multiple apps.

“Naturally, as we move into the AI ​​era, Apple is leaning more heavily into security as a core part of Apple Intelligence. But it doesn't escape people's attention that, like its competitors, it has had to offer cloud-based solutions for some capabilities,” said Ben Wood, chief analyst and CMO at consultancy CCS Insight. Registry.

“This comes with unique challenges. Apple has built its own private cloud for many functions, but the partnership with OpenAI on ChatGPT marks a new direction for Apple, which has historically been reluctant to partner on core technologies.”

The developer beta of iOS 18 is available to registered Apple developers starting today, with the public beta release planned for next month. Apple Intelligence will still be considered in beta testing when iOS 18, iPadOS 18 and macOS Sequoia are released to the public this fall.®



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *