Apple has described in a research paper how it trained its latest generative AI models using Google's neural network accelerators rather than, say, more fashionable Nvidia hardware.
Apple chooses Google over Nvidia for AI training
The paper [PDF], titled "Apple Intelligence Foundation Language Models," provides a deep dive into the iPhone titan's approach to LLMs, from training to inference.
These language models are the neural networks that transform questions and prompts into text and images, and drive the so-called Apple Intelligence features baked into Cupertino's operating system. They can do things like text summaries and suggested wording for messages.
While most AI organizations are clamoring for Nvidia GPUs, especially the H100 until Blackwell arrives—and may be looking at offerings from AMD, Intel, and others—when it comes to training machine learning systems, Apple decided to go with Google's Tensor Processing Unit (TPU) silicon. That's not entirely surprising, as the Mac Titan and Nvidia have been on bad terms for a few years now for various reasons, and it seems Cook & Co have little interest in patching things up to train Apple Foundation Models (AFMs).
