Why has Apple been unable to get AI right? Internal employees: It's like a sinking ship: "It has been sinking for a long time"

Recently, Bloomberg Businessweek published an in-depth report analyzing the reasons behind Apple's continued lag in the field of artificial intelligence and the significant setbacks in its AI strategy. Through interviews with several insiders at Apple, it revealed how executives, including Apple's AI chief John Giannandrea, have clashed with the company's long-established decision-making culture.

The article pointed out that as early as 2018, Apple recruited leading figure John Giannandrea from Google's AI sector, hoping he could lead Apple to breakthroughs in AI. However, seven years later, Apple's AI development has not gone as hoped, and the global popularity of large language models like ChatGPT has further intensified Apple's sense of crisis regarding AI.

Although Apple has launched new features like "Apple Intelligence," neither the timing of their release nor their actual performance has met expectations. Internally, some have likened the current situation to "a ship that is about to sink, 'it has been sinking for a long time'", profoundly reflecting the dilemmas and potential crises of the AI strategy.

Industry insiders have stated that the ongoing failures in artificial intelligence threaten plans ranging from the dominance of the iPhone to robotics and other future products.

"The Root of the Sinking Ship": Internal Structure and Decision-Making Dilemmas

The article stated that as the former head of Google's search and AI departments, John Giannandrea has found it difficult to integrate into Apple's core circle, where members have worked together for decades, managing the company like a family business. Like previous "parachute executives," he has also found it challenging to implement change. Within the company, he has had to face strong inherent cultural resistance, which directly impacts the implementation of Apple's AI strategy.

Moreover, Apple's software chief Craig Federighi has been reluctant to make large-scale investments in artificial intelligence. According to several colleagues, he does not believe that AI is a core function of personal computers or mobile devices. This attitude only began to change after the launch of OpenAI's ChatGPT at the end of 2022.

Giannandrea's alleged lack of urgency may stem from both philosophical considerations and his temperament. He has taken a conservative stance on the pace of AI development and has been skeptical about the value of chatbots. According to colleagues familiar with his thoughts, Giannandrea has argued internally that companies like OpenAI, Meta, and Google do not pose an urgent threat, insisting that the assistant users want is an interface for controlling devices, rather than a chatbot like ChatGPT.

Other executives share a cautious attitude similar to Federighi's. "In the world of artificial intelligence, you really don't know what the product is until you've completed the investment," said another senior executive. "That's not Apple's way of thinking. When Apple sits down to develop a product, it already knows what the end goal is."

Additionally, on a technical level, Apple has lagged behind competitors in acquiring key hardware resources like GPUs, resulting in slower training speeds for its AI models Finally, at the level of corporate culture, Apple's consistent style of "slow work produces fine products" seems out of place in the rapidly iterating AI field. Apple is accustomed to launching products only after the technology has matured, which stands in stark contrast to the fast-changing nature of AI.

"The Iconic Event of the Sinking Ship": The Difficult Birth of Apple Intelligence

The delayed launch of Apple Intelligence and the upgraded Siri has become a hallmark failure of Apple's AI strategy. In June 2024, Apple announced Apple Intelligence at the Worldwide Developers Conference, but many promised features were either delayed or performed poorly. The new version of Siri was originally scheduled for release in April 2025, but just weeks before the planned launch, Federighi personally tested it and found that many advertised features were not functioning properly.

Members of the Siri team indicated that the assessment by Robby Walker, the deputy to Vice President John Giannandrea, actually underestimated the severity of the issues. "There are now hundreds of bugs," one person said. "It's like whack-a-mole. You solve one problem, and three more pop up."

The main technical issue is that Apple had to split Siri's infrastructure in two: the old code supports traditional functions like setting alarms, while the new code supports requests that utilize personal data. This "temporary patchwork" approach was originally intended to bring new features to market quickly, but it backfired, leading to integration problems that ultimately delayed Siri's functionality.

Internal Apple employees revealed that the issues were primarily concentrated in the smooth integration of code, low morale among the development team, and instability in the testing environment, resulting in multiple features being "battle-scarred." The situation was likened to "a ship in a storm," with outdated architecture and technology routes being the main obstacles.

The Path Forward: Apple's AI Future

Many industry experts believe that if Apple cannot achieve breakthroughs in AI, the future ecosystem of smart devices may be "marginalized." "It seems that Apple's stage is gradually shrinking, falling behind competitors in smart assistants and automation, and relinquishing important technologies that may represent the future," pointed out an industry authority.

In the face of challenges, Apple is actively adjusting its strategy. The company has now created a new software architecture at its AI office in Zurich to replace the problem-ridden Siri hybrid system, which is a "monolithic model" built entirely on LLM engines, ultimately making Siri more conversational and better at synthesizing information.

Apple is also adjusting its AI branding strategy, preparing to separate the Apple Intelligence brand from Siri in marketing. This is the company's implicit acknowledgment that the poor reputation of its voice assistant may affect its AI messaging. Another change is that Apple will, in most cases, stop announcing features that are more than a few months away from official release.

Despite facing severe challenges, Keith Rabois, co-founder of Siri's original application, remains optimistic about an AI-driven Siri. "All the foundational model companies don't know what an assistant is, while Apple has been researching this concept since 2010," he said. He believes that all Apple needs to do is make Siri smarter They still have the buttons and the brand, and if they perform a 'brain transplant' for Siri, they have a chance to become the preferred assistant