Apple’s foray into artificial intelligence has faced significant challenges since the debut of its initial features last year. A comprehensive report from Bloomberg’s Mark Gurman sheds light on the company’s struggles and its ongoing efforts to revamp its AI capabilities, focusing particularly on a complete redesign of Siri.
Gurman previously indicated that Apple is developing what it refers to internally as ‘LLM Siri’—a generative AI rebuild of its digital assistant. The previous hybrid approach, which sought to enhance Siri using existing frameworks, has not yielded the desired results. Gurman outlines several key factors contributing to the current state of Apple Intelligence:
- Resistance to major AI investments. Craig Federighi, Apple’s software chief, has shown hesitation towards making hefty investments in AI technologies. An unnamed executive noted that in the realm of AI, clarity on product direction typically emerges post-investment, leading to delays in acquiring the necessary computing resources to remain competitive.
- Late entry into the AI scene. According to another source, Apple Intelligence lacked a concrete vision prior to the launch of ChatGPT in late 2022.
- Misjudgment of consumer preferences. John Giannandrea, Apple’s head of AI, reportedly believed there was little demand for AI chatbots, suggesting that customers preferred the option to disable such tools.
- Integration issues between old and new technologies. Apple’s strategy of integrating generative AI features with the traditional Siri has not proven effective, leading to a frustrating cycle of fixing one problem only to encounter several new ones, according to internal sources.
- Cultural disconnect within leadership. Giannandrea, having been a notable external hire, reportedly did not possess the assertiveness often found in Apple’s upper management. His lack of urgency in securing substantial funding raised concerns among employees, who felt he did not adequately address competitive threats from major AI firms.
- Over-ambitious marketing strategies. The company’s AI marketing efforts made bold claims about enhanced features for Siri and Apple Intelligence, despite those features being unready for deployment, resulting in significant delays.
In response to these hurdles, Apple is rethinking its strategic approach. A comprehensive revamp of Siri is underway, with efforts directed towards creating a new framework rather than merely adapting the old system. Gurman reports that Apple’s AI team in Zurich is developing a new architecture based on an LLM engine, designed to enhance conversational capabilities and better synthesize information.
Additionally, Apple is exploring the utilization of iPhones and differential privacy to refine its synthetic data. This process involves analyzing non-personally identifiable data from user communications to improve AI training, all while maintaining user privacy by keeping real data on-device. The company is also considering enhancements to Siri’s capabilities by enabling the LLM version to access and compile information from the web, similar to existing AI search tools—one of which, Perplexity, has been approached by Apple for potential collaboration.
As these changes unfold, John Giannandrea is reportedly stepping back from his direct involvement in product development, including projects related to Siri and robotics. Talks within Apple suggest a gradual transition towards retirement for him, though there are concerns about the potential fallout if members of his team were to leave alongside him. Nevertheless, it appears Giannandrea intends to remain with the company, expressing relief that the challenges surrounding Siri will now belong to another leadership team.
