Would You like a feature Interview?
All Interviews are 100% FREE of Charge
Apple CEO Tim Cook (left), Senior Vice President of Machine Learning and AI Strategy John Giannandrea (center), and Senior Vice President of Software Engineering Craig Federighi (right) speak during Apple’s annual Worldwide Developers Conference in Cupertino, California on June 10, 2024.
Nick Cooley | AFP | Getty Images
apple Apple on Monday announced its full embrace of artificial intelligence, explaining the features and rationale behind Apple Intelligence, its new AI software suite.
But Apple’s Worldwide Developers Conference launch event, with a panel discussion on Monday afternoon, was carefully planned to distinguish the iPhone maker from current AI leaders such as Microsoft and Google.
During a panel, software chief Craig Federighi and AI chief John Giannandrea said Apple approaches technology differently than its Silicon Valley rivals: Unlike companies building AI for a wide range of products, Apple is focused only on the devices it sells and the personal data its AI can use.
Apple has revealed a more limited approach that eschews future-focused thinking about the technology’s potential, instead emphasizing small tasks that can be completed right now without draining battery life.
“We believe the role of AI is not to replace users, but to empower them,” Federighi said.
Apple’s AI may be the first thing its 2 billion users are exposed to, if its AI capabilities are preferred over cloud-based competitors. Microsoft or GoogleThis could change how multi-billion-dollar-a-year AI infrastructure is built and reshape the direction of products that use the technology.
Many of the AI ​​developments that have captured the interest of investors and tech professionals are focused on building or securing powerful supercomputers with the following capabilities: NVIDIA The chips are then used to develop more power-hungry AI models. In this scenario, users would access the AI ​​software by communicating over the web with an equally powered server.
Apple’s AI lives primarily on device
Apple Intelligence was announced at Apple’s Worldwide Developers Conference in Cupertino, California on June 10, 2024.
Source: Apple Inc.
Apple’s vision for AI is not one big model, but many smaller models that don’t require the same amount of computing power or memory, running on Apple devices and on the chips themselves. If the AI ​​on the phone can’t do it, Apple, or apps that use Apple’s tools, will go to the cloud to access the larger AI model. For example, Apple has partnered with OpenAI to give users access to ChatGPT when Siri can’t answer. These features only work if the user allows it.
Apple executives aren’t calling this a one- or multi-model strategy, instead simply calling it “Apple Intelligence.”
“We think the right approach to this is to have different models and different sizes for different use cases,” Giannandrea said.
Giannandrea said the company worked on creating a 3 billion parameter model as part of Apple Intelligence. By comparison, ChatGPT’s 2020 GPT-3 model is much larger, with 175 billion parameters. The more parameters, the more memory and computing power required to run the model.
Apple’s approach is faster than cloud-based options and has privacy benefits, but it could run into problems if the models are too small to do anything about. Apple believes that through a user’s iPhone, its AI can leverage personal data about schedules, location and user behavior. One example Federighi gave was that his phone knows who his daughter is.
Apple also said that rather than giving users an unlimited number of chatbot interfaces, it is allowing its smaller models to work only on tasks they excel at.
“There’s an important extra step, which is we’re not taking this young person and telling him to fly a plane,” Federighi said.
Many of the AI ​​features Apple unveiled on Monday are similar to products it has already announced this year. Apple’s AI can summarize and rewrite documents, generate tiny images and translate conversations in real time. One notable feature will allow users to use AI to generate new emojis without an internet connection. The new features are expected to be released in beta this fall.
Apple’s approach to privacy
Private Cloud Computing was announced at Apple’s Worldwide Developers Conference on June 10, 2024 in Cupertino, California.
Source: Apple Inc.
Privacy will be a challenge for Apple as it embraces AI, as the company has long used it as one of its primary marketing tools, stressing that its business model doesn’t require ad targeting and that it puts users’ interests first, rather than data brokers and spammers.
Other AI companies collect and store user data to improve their software, but that doesn’t fit with Apple’s current privacy policy. Much of Apple’s presentation on Monday was devoted to the steps the company has taken to avoid the impression that it is siphoning user data to improve its AI.
“We’re not going to take that data and send it to a cloud somewhere,” Giannandrea said, “because we want everything to be very private, whether it’s running locally or on a cloud computing service. That’s how we want to make sure we’re able to use the user’s most personal data.”
Apple did not elaborate on what data was used to train the AI ​​models, other than to use files collected from the public web, in addition to licensed data such as news archives and stock photos.
Apple said it developed its own servers, called the Apple Private Cloud, using Apple chips to ensure that user data sent back to the AI ​​servers isn’t stored or reused — a notable move for a secretive company that allows third parties to inspect its software but typically doesn’t provide information about its infrastructure.
“Even if a company promises, ‘You know, we’re not going to do anything with the data,’ there’s no way to verify that,” Federighi said, explaining why Apple allows testing of its AI server software.
More AI on the way
ChatGPT’s integration with Apple iOS 18 was announced at Apple’s Worldwide Developers Conference in Cupertino, California on June 10, 2024.
Source: Apple Inc.
Apple executives have at times seemed to downplay that this is a major shift in the company’s AI strategy, saying it’s just an extension of the machine learning work the company has already done for photo editing, text transcription, and even to include a dedicated block of AI on its chips.
“Some people have suddenly started claiming that there’s some new category out there, but these are things we’ve been shipping for a long time,” Federighi said.
But Apple isn’t betting everything on a single approach. The company is integrating ChatGPT into its operating system, allowing users to try OpenAI’s models for free and giving users access to more powerful, larger AI models. However, OpenAI’s ChatGPT will be marked in Apple’s software to notify users that their data will be sent to OpenAI servers running on Microsoft’s cloud. Answers will also be marked as having been generated by ChatGPT, just in case they go off track.
He said Apple may offer different models in the future, suggesting that Apple Intelligence is not the only AI system the company expects customers to use. Federighi said some customers in the future might want to embed a medical AI system or a legal AI model in an Apple product, for example. Or maybe one of Google’s models.
“In the future, we’re looking forward to integrating with models like Google Gemini, for example. We don’t have anything to announce at this time,” Federighi said. “But that’s the direction we’re heading.”