Apple Reboots Siri with New AI Architecture to Compete in Generative AI Race

Apple is working to revitalize its struggling AI initiatives, particularly focusing on revamping Siri. A recent report by Bloomberg’s Mark Gurman delves into the challenges and the company's renewed efforts. Here’s a condensed overview: Apple's AI strategy has faltered since the introduction of new features last year, notably in its digital assistant, Siri. Key issues identified include reluctance to invest heavily in AI, a late start compared to competitors like ChatGPT, and an internal misalignment between old and new AI technologies. John Giannandrea, Apple’s AI chief since 2018, struggled to secure significant funding and was hesitant to fully embrace AI chatbots, believing customers preferred disabling such tools. Craig Federighi, Apple’s software chief, was also cautious about making substantial AI investments without a clear end goal, leading to a lack of necessary hardware and delayed progress. Marketing overshot the mark, promising advanced features like a more context-aware Siri and Apple Intelligence before they were fully developed, resulting in delays and disappointment. To address these issues, Apple is now embarked on a major overhaul of Siri, aiming to develop a completely new assistant based on large language models (LLMs). The AI team in Zurich is leading the charge, working on a new architecture that will leverage LLMs to make Siri more conversational and adept at synthesizing information. This involves a shift from the current hybrid approach, where generative AI was bolted onto the existing Siri, to a ground-up rebuild. The new architecture aims to resolve the persistent "whack-a-mole" issue, where fixing one problem led to the emergence of several others. Moreover, Apple intends to use iPhones and differential privacy to enhance its AI data synthesis. By comparing synthetic data with user-generated content on-device, Apple can improve its AI training without compromising user privacy. The company is also considering allowing the LLM version of Siri to access the web, much like Perplexity, to gather and synthesize data from various sources. This could transform Siri into a powerful web search tool, similar to those found in AI-driven platforms. Despite these ambitious plans, Giannandrea has been removed from product development, Siri, and robotics projects since the spring. There are rumors that he may be on a path to retirement, although he remains with Apple, reportedly relieved that Siri is no longer under his purview. The shift comes as Apple realizes the need for a more forceful and proactive approach to AI, recognizing the urgent threat posed by competitors like OpenAI and Google. Industry insiders view Apple’s new strategy positively, seeing potential in the company’s focus on privacy-preserving AI techniques and the integration of LLMs for a more natural and intelligent assistant. However, they note that Apple’s success will depend on its ability to make timely and substantial investments in AI infrastructure and talent. Apple’s strong position in hardware and user base provides a solid foundation for these efforts, but the company must also adapt its traditionally conservative approach to keep pace with rapid advancements in the field. Apple continues to be a leader in the tech industry, known for its innovation and rigorous standards. With the new AI initiatives, the company is demonstrating its commitment to staying relevant and competitive in the ever-evolving landscape of digital assistants and broader AI applications.