Apple is developing two major artificial intelligence initiatives, codenamed Linwood and Glenwood, “as part of a sweeping plan to make far more capable and competitive in the generative AI era,” reports AppleInsider.
Read the article for all the details, but here are two highlights:
° Linwood is Apple’s in-house effort to rebuild the Siri “brain” around large language models developed by the Apple Foundation Models team. These models are designed to understand natural language, manage complex queries, and retain conversational context.
° While Linwood focuses on internal innovation, Glenwood explores a different route. Apple is testing Anthropic’s Claude to power Siri’s core intelligence and has experimented with tools like ChatGPT and Google Gemini during development. The projects represent a notable shift in Apple’s traditionally insular approach. The company has historically relied almost entirely on in-house technology.
In April it was reported by The Information that Craig Federighi, Apple’s senior vice president of Software Engineering, told Siri engineers to do “whatever it takes to build the best AI features,” even if that means using open-source models from other companies in its software products as opposed to Apple’s own models.
From the report: Federighi has already shaken things up. In a departure from previous policy, he has instructed Siri’s machine-learning engineers to do whatever it takes to build the best AI features, even if it means using open-source models from other companies in its software products as opposed to Apple’s own models, according to a person familiar with the matter.
The Information noted that, until now, Apple engineers could only use third-party LLMs to benchmark them against their own in-house models during testing. Under Federighi’s leadership, it seems that all open-source LLMs can be used by Apple engineers.
A LLM (large language model) is a type of machine learning model designed for natural language processing tasks such as language generation. They’re used in a wide range of applications, including chatbots, content creation, translation, and more. Some well-known examples include GPT (Generative Pre-trained Transformer) models, which are used in applications like ChatGPT.
