OpenAI and Apple's partnership, explained

Why did Apple partner with OpenAI?

There are numerous reasons why Apple decided to partner with OpenAI, including the following:

• OpenAI is an industry leader. OpenAI, one of the leading vendors in the generative AI era, is well known for ChatGPT, a brand that Apple users will likely recognize and trust.
• The partnership provides an understandable story for investors. It provides a solid AI angle for Apple with a proven leader. The partnership could help reassure investors that Apple is keeping up with AI trends and remaining competitive.
• Users will have enhanced AI capabilities. Integrating ChatGPT allows Apple to offer advanced AI features to its users that it might not yet have on its own.
• Siri will have a supplement. A key integration point for the partnership is with Apple's Siri personal assistant. ChatGPT will be used when Siri needs additional assistance answering queries, providing an expanded range of tasks and questions that Apple's digital assistant can handle.
• It provides broader internet knowledge. The core models inside Apple Intelligence are primarily focused on personal assistance with user data, while ChatGPT provides broader information about history, current events and general knowledge.
• Apple will balance user privacy and advanced AI. The partnership is structured so that Apple can maintain its focus on user privacy while still providing cutting-edge AI capabilities.
• It gives Apple a competitive advantage. By offering ChatGPT integration, Apple can differentiate its products from competitors and potentially attract more users to its ecosystem.
• There are mutual benefits for both companies. While Apple gains access to advanced AI technology, OpenAI benefits from exposure to millions of iPhone users.

How is Apple Intelligence different from ChatGPT?

Apple Intelligence is different from ChatGPT in numerous ways. While Apple has partnered with OpenAI to integrate ChatGPT into some of its services, ChatGPT is separate from the core Apple Intelligence features.

How will Apple use data for the ChatGPT applications?

Apple is taking a cautious, privacy-focused approach to data usage for applications with ChatGPT. This includes practices such as the following:

• Opt-in consent. ChatGPT access is not automatic. Apple's users will be required to give their consent each time an application wants to use ChatGPT to answer a question. This repeated consent prompt is a key privacy measure.
• Limited data sharing. Apple pledged not to share any personal user information with OpenAI. Inquiries made through ChatGPT will not be linked to an Apple user's account. Apple has stated that OpenAI will not store user requests sent through the ChatGPT integration, nor will OpenAI use information from Apple users for model training.
• Data transparency. Apple is trying to be clear about what is being done with user data, especially concerning the multimodal elements of OpenAI's GPT-4o model that can access device cameras and screens.
• IP address protection. As an additional privacy measure, Apple is obscuring users' IP addresses when interacting with ChatGPT through Apple platforms.

How did Apple train its AI models?

Apple used a variety of techniques to train its own AI models. Apple Intelligence includes multiple models, such as both text- and image-based foundation models. Many models were developed and trained using the Apple AXLearn framework, including a 3 billion-parameter large language model that provides summarization and text generation capabilities.

The training used a combination of data parallelism, tensor parallelism, sequence parallelism and fully sharded data parallel. Apple uses various training hardware and cloud platforms, including tensor processing units and both cloud and on-premises GPUs.

For the training data, Apple used a combination of publicly available and licensed data. Apple emphasizes that it never uses private personal data or user interactions when training its foundation models. Going a step further, Apple filters out profanity and other low-quality content from its training data.

How will ChatGPT work with Apple platforms?

The OpenAI partnership will bring ChatGPT to Apple's iOS 18, iPadOS 18 and macOS Sequoia operating systems.

Users will have the option to access ChatGPT through Siri, primarily as a supplement when Siri needs additional assistance answering queries. ChatGPT will also be an available option in Apple's systemwide writing tools on iPhone, iPad and Mac devices. In addition, Apple users will get access to ChatGPT's image generation capabilities, which are powered by OpenAI's Dall-E 3 model.

Each time Siri or an Apple writing tool wants to use ChatGPT, the system will prompt the user for permission. Users will not need to switch out of the app to access ChatGPT, as the experience and results are all integrated.

Apple users will not need to create a separate account with OpenAI to use the basic ChatGPT features. However, ChatGPT Plus subscribers can connect their account to access additional paid features from OpenAI within the Apple ecosystem.

Are there any industry concerns?

The partnership between Apple and OpenAI has raised numerous industry concerns, which include the following:

• Privacy and data security. Regulators and former OpenAI employees have raised questions about the extent of data sharing between Apple and OpenAI, and how user privacy will be protected. For example, Elon Musk voiced strong concerns about potential security violations if OpenAI is integrated at the operating system level on Apple devices.
• Transparency and oversight. Regulators in Europe and Japan are calling for external oversight to ensure the enforcement of safety and security measures. There are concerns about how Apple will verify that OpenAI is deleting user data sent to its servers.
• Internal issues. Executive departures and conflicts at OpenAI related to safety concerns have raised questions about the company's priorities. There are worries about the rapid deployment of AI technologies without adequate safeguards.
• Competition and antitrust issues. Industry experts have concerns about how this collaboration might affect competition in the AI space. In addition, the partnership could attract regulatory scrutiny, given Apple's ongoing antitrust lawsuit with the U.S. Department of Justice.
• Device functionality. There are concerns about how the integration of ChatGPT might potentially affect Apple devices' performance, including speed.

What could this partnership mean for the future of generative AI?

The Apple partnership with OpenAI could have several implications for the future of generative AI, including the following:

• Wider adoption. The integration of ChatGPT into Apple devices could bring generative AI technology to hundreds of millions of new users, potentially accelerating mainstream adoption.
• Increased scrutiny. As generative AI becomes more widespread through partnerships like this, it's likely to face greater scrutiny regarding safety, privacy and ethical concerns.
• Focus on privacy. Apple's approach to implementing privacy controls and safeguards with ChatGPT could set a precedent for how other companies integrate third-party AI technologies.
• Added competition in AI development. This move could encourage other tech giants to form similar partnerships or accelerate their own AI development efforts to stay competitive.
• Increased AI investment. The positive market reaction to Apple's partnership could encourage further investment in generative AI across the tech industry.

How the OpenAI and Apple partnership plays in China

China's strict regulations on generative AI have resulted in ChatGPT being banned in the country.

This ban poses a significant problem for Apple, as China is its second-largest market, and iPhone sales are under pressure from domestic rivals. The lack of full AI functionality could further deter Chinese consumers.

To address this issue, Apple is seeking a Chinese AI company to partner with before the official launch of iOS 18 in September.

Sean Michael Kerner is an IT consultant, technology enthusiast and tinkerer. He has pulled Token Ring, configured NetWare and been known to compile his own Linux kernel. He consults with industry and media organizations on technology issues.