Apple's new AI is made in Google data centers
What specific AI features is Apple introducing that require the use of Google's silicon for training?
Apple is introducing several AI features that require the use of Google's silicon for training. These features include:
-
Apple Intelligence: A personal intelligence system integrated deeply into iOS 18, iPadOS 18, and macOS Sequoia1. It comprises multiple highly-capable generative models specialized for users' everyday tasks and can adapt on the fly for current activities.
-
On-device language model: A ~3 billion parameter on-device language model that is fine-tuned for user experiences such as writing and refining text, prioritizing and summarizing notifications, and taking in-app actions to simplify interactions across apps.
-
Server-based language model: A larger server-based language model available with Private Cloud Compute and running on Apple silicon servers1. This model handles more complex tasks and is trained using Google's Tensor Processing Units (TPUs).
-
Adapter models: Small neural network modules that can be plugged into various layers of the pre-trained model to fine-tune the models for specific tasks. These adapter models are trained using Google's TPUs and both cloud and on-premise GPUs3.
-
Image Playground: A feature that allows users to generate images using AI in Messages, Photos, Notes, Keynote, Freeform, Pages, and third-party apps5. It supports Animation, Illustration, and Sketch styles and is trained using Google's TPUs and both cloud and on-premise GPUs.
-
Custom emoji: Apple is developing software that can create custom emojis on the fly, based on what users are texting. This feature is also trained using Google's TPUs and both cloud and on-premise GPUs.
Overall, these AI features are designed to help users communicate, work, express themselves, and get things done across their Apple products. They are built with privacy in mind and leverage a combination of on-device processing and cloud-based AI training to deliver a powerful and personalized experience.
How do the partnerships between tech giants like Apple and Google reflect the current dynamics in the AI industry?
The partnerships between tech giants like Apple and Google reflect the current dynamics in the AI industry, highlighting the importance of collaboration and leveraging each other's strengths to advance AI technology. These partnerships also signify the highly competitive nature of the AI market, with companies working together to stay ahead of the competition and meet consumer demands.
Apple's partnership with OpenAI, which will put ChatGPT on millions of iPhones, showcases the desire for tech companies to integrate cutting-edge AI features into their products. Additionally, Apple's reliance on Google's data centers to train its AI models demonstrates that even rival companies recognize the benefits of collaboration in the AI space.
Google's close call in delivering the required Tensor Processing Units (TPUs) for Apple's AI training highlights the increased demand for cloud computing resources and AI-specific chips. Companies like Microsoft, Google, and Amazon have a competitive advantage, having already invested in building massive data centers for AI training.
The potential partnership between Google and Apple to bring Google's Gemini AI to iOS devices further emphasizes the importance of collaboration in the AI industry. Both companies have much to gain from such a partnership: Google could become the preferred generative AI service on both Apple and Android phones, while Apple could integrate AI more quickly and potentially offer users a choice of AI chatbot providers16.
These partnerships and collaborations demonstrate that even the largest tech companies recognize the need to work together to advance AI technology and remain competitive in the rapidly evolving AI landscape.
What led to the technical crisis at Google following Apple's request for more TPUs?
Apple's request for additional Tensor Processing Units (TPUs) to train its AI models led to a technical crisis at Google. This created a scramble inside Google in April, as there were technical issues that could have prevented them from delivering what Apple wanted on time. The situation was treated as an "OMG," a Google term for one-off urgent incidents that don't quite warrant a code red. A war room was convened inside Google to address the issue, and the team was able to deliver for Apple after a few very long days. The crisis underscores the complex relationships between tech giants like Apple and Google, who are both partners and competitors in the AI race.