According to renowned tech journalist Mark Gurman, Apple, a tech behemoth, is confronting a pivotal juncture in its AI strategy. Internal evaluations suggest abandoning heavily invested self-developed large language models in favor of collaborating with AI leaders OpenAI and Anthropic. This shift aims to infuse Siri, Apple's voice assistant, with more robust intelligence capabilities.

This potential strategic pivot casts a pall over Apple's ambitious "Apple Intelligence" vision. At last year's WWDC, Apple outlined a three-tier AI approach integrating on-device intelligence, cloud-based self-developed models, and external large models, with privacy as its cornerstone. However, a year later, promised AI functionalities—especially Siri's intelligent enhancements—have frequently been postponed, with implemented models criticized for being inferior to competitors, fostering user and market disappointment. Once an AI trailblazer, Apple now stumbles in the realm of large models, with internal constraints and ecosystem limitations compelling Tim Cook to make tough decisions.

01. Promises are Easy, Delivery is Hard: Apple's AI Frequently Misses Deadlines

At WWDC 2024, Tim Cook was upbeat. Apple redefined AI as "Apple Intelligence," proposing a three-tier architecture of "small on-device models + cloud-based self-developed models + external large models," with user privacy as a core selling point. Specific functionalities included notification summaries, speech-to-text, writing tools, image generation, and smart photo removal to enhance user experience.

Siri's upgrade was particularly noteworthy: Apple introduced a text input mode and planned to equip it with continuous conversation, screen perception, and enhanced intelligence. For instance, at last year's conference, Siri combined email content and flight status to provide arrival times when asked, "When does Mom's flight land?"

However, at WWDC 25, Apple unveiled "Liquid Glass," a new user interface featuring transparent visual elements, a new game aggregation app, and expanded features for CarPlay and AirPods. The AI response was muted. Apple announced it would open its basic AI models to developers, allowing offline usage. However, reports suggest these models are simpler than those of OpenAI, Google, and ChatGPT.

Some Apple fans expressed, "Although we didn't have high hopes, we're still disappointed." Others believed Apple's leadership was too conservative, failing to introduce groundbreaking AI functionalities or significant progress during the conference.

Furthermore, planned AI-assisted Siri upgrades were incomplete. Federighi stated, "We need more time to meet the company's high standards."

02. From Leading to Lagging: Even a Skilled Chef Needs Ingredients

Apple's Siri once dominated globally, showcasing smart voice assistants' potential. Apple was also an early adopter of integrating machine learning and AI into system functionalities, with features like automatic portrait categorization in photo albums.

Apple extensively uses machine learning in image algorithms, predictive input, and Face ID to optimize user experience. The early adoption of AI in iOS, alongside A-series chip power, played a crucial role in iPhone's success.

In 2023, as large AI models gained public attention, Apple launched "Ajax," an AI large model project, publishing related papers and results the following year. Many believed Apple would pioneer AI in 2024.

However, Apple Intelligence's launch was severely delayed. Even in North America, not all WWDC 2024-demonstrated features were rolled out. Post-launch, users found the experience fell short of Apple's promises, exacerbating market and user distrust in Apple's AI projects.

Apple's lagging behind is due to multiple factors. In 2018, Apple poached John Giannandrea from Google to head its AI department, aiming to swiftly address AI shortcomings. Giannandrea, formerly Google's second-in-command, integrated AI into Google's core products, giving Google a lasting industry lead.

Despite resources and determination, Apple failed to overtake the competition. In Bloomberg's "Why Apple Still Hasn't Cracked AI," Mark Gurman revealed insiders' perspectives on Apple's AI lag.

Firstly, executive consensus is lacking. Giannandrea initiated reforms, including large investments in machine learning model training, Siri department restructuring, and cutting low-usage functionalities. However, his reforms faced resistance from Federighi, who was reluctant to invest heavily in AI, believing it's not core to personal computers or mobile devices. Without support, Giannandrea's AI progress was slow.

This internal friction delayed key decisions, such as the Siri large model upgrade project being postponed thrice due to architecture conflicts. Additionally, core AI talents were poached by OpenAI and Anthropic, with 11 executives leaving in 2023 alone.

Secondly, Apple's closed ecosystem hinders innovation. While Google opened its Gemini model to attract global developers, Apple's Core ML framework restricts third-party innovation. More critically, privacy protection and data hunger contradict—differential privacy leads to a lack of training data, resulting in inferior model accuracy.

03. Apple Still Holds Its Own Advantages

Despite challenges, Apple retains core AI advantages, such as its unparalleled user scale and depth.

Firstly, Apple boasts over 2 billion active devices globally, providing high-frequency, daily, and immersive usage scenarios.

Secondly, these devices are deeply integrated into users' lives, serving as hubs for communication, work, entertainment, health, payments, and smart home control. For AI to thrive, it must penetrate these scenarios, making Apple devices ideal carriers.

Thirdly, Apple users typically have high spending power and technological acceptance, making them ideal for AI services (especially paid ones).

For AI companies aiming to maximize user reach, Apple's App Store and operating system platforms are indispensable. They need to develop iOS/macOS apps or integrate with Apple services, giving Apple strong bargaining power in cooperation and platform rule-making.

04. Conclusion: Apple's Main Competitor is Itself

Apple's AI journey reflects tech giants' transformation dilemmas. From Siri's pioneering glory to frequent AI delays, Apple's struggles reveal a harsh reality: irreconcilable contradictions between technological ambitions, internal constraints, closed ecosystems, and open innovation needs. Executives' AI strategic value disagreements lead to insufficient resource investment and delayed decisions. Core talents are poached, weakening R&D foundations. Privacy protection and closed ecosystems, once prized, now hinder data-driven AI innovation, causing self-developed models to lag behind in accuracy.

However, Apple has options. Its true strength lies in controlling a vast ecosystem of over 2 billion active devices deeply integrated into users' lives. This reach gives it a unique platform advantage in the AI race. Whether Anthropic or OpenAI, any AI company aiming to maximize user scale and scenario penetration cannot bypass Apple. This advantage emboldens Apple to consider abandoning solo efforts and embracing external collaboration.

Apple's AI journey stands at a critical juncture. Abandoning some self-development fantasies and embracing top-tier external models may be the most pragmatic choice to resolve internal difficulties and leverage ecological advantages. This isn't a compromise but a strategic integration, leveraging platform power to translate global AI innovations into user experiences. Navigating this internal challenge will determine Apple's resurgence in the intelligent era.

- End -