In the world of AI, Apple appears to be hitting a significant roadblock as its digital assistant, Siri, struggles to keep pace with technological advancements. A revealing report by Bloomberg's Mark Gurman sheds light on the turmoil within Apple Intelligence, detailing the company's efforts to revive Siri with a fresh twist.

Apple is now focusing on what insiders are calling 'LLM Siri,' a transformative generative AI iteration of Siri. This comes after the previous strategy of combining the old Siri with generative features failed to produce satisfactory results.

According to Gurman, several factors contributed to the sluggish progress. Apple’s software chief, Craig Federighi, was initially hesitant to pour resources into AI initiatives, often preferring clear objectives before investing. This caution meant the company fell behind in acquiring essential technology, like powerful GPUs, which competitors quickly secured.

A late start meant that by the time ChatGPT made waves in late 2022, Apple hadn’t even conceptualized its AI strategy. AI chief John Giannandrea believed that users were uninterested in chatbots, leading to a slower rollout of innovative AI features.

The attempt to integrate new AI capabilities with the outdated Siri resulted in a failure to address fundamental flaws. An Apple employee likened the situation to 'whack-a-mole'—tackling one issue only for multiple new problems to arise.

Giannandrea’s position as an outsider within Apple’s leadership may have hampered progress. Lacking a forceful presence, he reportedly didn't secure the substantial funding needed for ambitious projects, while underestimating the urgency of rival companies like OpenAI and Google.

Marketing also played a role in the missteps. Apple’s promotional materials suggested enhancements to Siri that were not yet ready for launch, leading to promised features being delayed.

Now, Apple has shifted its strategy. The focus is on a complete overhaul of Siri rather than patching up the old model. The goal is to create a new architecture based on a LLM (Large Language Model) engine that will enable Siri to engage in more natural conversation and synthesize data more effectively.

Apple's AI team in Zurich is at the forefront of this transformation, utilizing advanced techniques like differential privacy to improve data synthesis directly on users’ devices. This would allow Siri to gather information from various sources online, similar to web search tools like Perplexity, which Apple has considered partnering with for AI search in Safari.

Interestingly, Giannandrea has stepped back from direct involvement in developing products, robotics, and Siri. Although discussions about his retirement path are underway, he’s reportedly content to watch from the sidelines, believing that the future of Siri now lies in the hands of others.