Apple's "AI Ambitions" Hit Again! AI Large Model Leader Poached by Meta for Huge Sum

According to the Zhitong Finance APP, Mark Gurman, an Apple product leaker who has accurately revealed iPhone update details multiple times, stated in his latest post that, based on information from insiders, after poaching several core researchers from OpenAI, Meta Platforms, led by Mark Zuckerberg, has now completely extended its "money-spending recruitment hand" towards Apple.

Gurman indicated that insiders revealed that the top executive responsible for the development and deployment of artificial intelligence large models at Apple Inc. (AAPL.US) has switched to Facebook's parent company Meta Platforms (META.US). This undoubtedly deals a heavy blow to Apple's more personalized Siri AI voice assistant, which has been delayed by a year, as well as to Apple's already struggling development pace in the entire artificial intelligence field. Prior to the poaching of this AI large model executive by Meta, several executives related to AI and machine learning business had already announced their departure from Apple.

Insiders disclosed that one of the world's most outstanding AI engineers—Ruoming Pang, head of Apple's foundation models team (Apple foundation models, abbreviated as AFM), who is in charge of Apple's artificial intelligence large model project, is about to leave Apple.

It is understood that Ruoming Pang joined Apple in 2021 from Google's parent company Alphabet Inc., causing a sensation in Silicon Valley at the time. Professionals in the AI field generally believed that Pang's arrival would significantly advance Apple's planning and layout in artificial intelligence. After Apple announced in 2024 that it would launch a more intelligent Apple Intelligence feature and Siri AI voice assistant on the iPhone, Pang was regarded as a key figure in helping Apple ultimately achieve its "artificial intelligence ambitions."

Spending heavily to poach top AI talent! Meta's crazy spending on AI recruitment continues unabated

Insiders stated that Ruoming Pang will become the latest heavyweight member of Meta's newly established "Super Intelligence" team. As the personnel announcement has not yet been made, insiders requested anonymity.

Reportedly, to recruit Ruoming Pang, Meta offered a compensation package with an annual salary reaching tens of millions of dollars. Meta CEO Mark Zuckerberg has recently been aggressively hiring the world's top AI talent, having recruited Scale AI founder Alexandr Wang, top AI startup founders Daniel Gross and former GitHub CEO Nat Friedman, as well as leading figures in the AI field such as Jiahui Yu, who led the OpenAI o3, o4-mini, and GPT-4.1 projects, and Hongyu Ren, creator of the o3-mini and o1-mini models.

Insiders noted that Meta also announced on Monday local time that it had hired researcher Yuanzhi Li from OpenAI and Anton Bakhtin, who previously participated in the development of the Claude large model project at Anthropic, known as an "OpenAI rival." Meta declined to comment. Apple, OpenAI, Anthropic, and Pang himself also did not respond to requests for comments.

In the face of competitors like OpenAI and Google, Zuckerberg has made the development of AI technology Meta's top priority. He is personally involved in recruiting the top talents in the AI development field, hosting potential AI candidates multiple times at his mansions in Silicon Valley and Lake Tahoe, and personally engaging with them.

At the end of June, Zuckerberg reorganized the company's AI team, focusing on "superintelligence"—AI technology that can perform tasks at or even above human IQ and learning capabilities. Meta's management previously stated that this year it would invest billions of dollars in AI-related fields, primarily for the construction and expansion of data centers, as well as spending heavily on AI chip and computing infrastructure.

The poaching of the leader responsible for Apple's AI models occurred shortly after Meta spent nearly $15 billion to acquire a 49% stake in Scale AI, and hired the company's CEO, Alexandr Wang, to lead Meta's new "superintelligence" division. This deal could reshape the competitive landscape in the AI data labeling field and highlight Meta's strategy of accelerating its AI layout through massive investments in the AI application software race.

For Meta, in the context of global companies vying for dominance in AI applications, Scale AI may help Meta create "killer" AI applications. As a leader in global data labeling and model evaluation, Scale AI holds the "data production materials" of the AI era. Scale AI can provide an industry-leading data labeling and model evaluation platform, while Meta is eager to commercialize the Llama series of large models on a large scale and deeply embed Meta AI into its social, advertising, and hardware ecosystem. Scale AI can be seen as the last key piece of Meta's "AI moat": the integration of computing power (NVIDIA AI GPUs + self-developed ASICs + large-scale data centers covering the globe), large models (Llama family), and data (Scale AI).

Under Zuckerberg's leadership, Meta seems very determined to adopt a massive investment model of "spending money to create miracles" to lay out artificial intelligence technology. Since the beginning of this year, Facebook and its parent company Meta have actively negotiated heavyweight mergers and acquisitions related to AI, and have poached top talents from competitors like Google's parent company Alphabet (GOOGL.US) and OpenAI to form a "superintelligence team."

The unprecedented AI acquisition frenzy initiated by Meta also reminds investors of the company's early aggressive practices (when it was still called "Facebook") in rapidly acquiring Instagram and WhatsApp during the global social media wave to suppress competition. The acquisitions of Instagram and WhatsApp over a decade ago have already proven Meta's keen investment insight Apple loses another AI general! More AI backbone will leave Apple in the future

At Apple, Ruoming Pang led a research team of about 100 people, responsible for the development of the company's large language models, which support Apple Intelligence and other device-side AI functions. In June of this year, Apple announced that it would open these models to third-party developers for the first time, bringing a variety of AI applications to the iPhone and iPad.

However, insiders revealed that the Apple Foundation Model team (AFM team) has recently faced scrutiny from new management, which is exploring the use of third-party AI large models from OpenAI or Anthropic to support the new version of the Siri AI voice assistant. These internal discussions have significantly weakened the morale of the AFM team in recent weeks.

Gurman stated that while Apple is considering empowering the new Siri AI voice assistant development phase with third-party solutions, it is also developing a new version of Siri based on the models from Pang's team. These models also drive Genmoji, as well as priority notifications, emails, and web summaries among other Apple Intelligence features.

In the vision of Apple fans for the Siri AI update iteration, with the support of cloud and device-side AI large models, the positioning of Apple Siri may no longer be a clumsy formal voice assistant. By combining cloud AI computing resources and device-side generative artificial intelligence capabilities, Apple iPhone models are expected to achieve a "personal AI assistant" that better meets individual user needs, similar to the "omni-functional AI companion" in the movie "HER." Apple has stated that the updated Siri voice assistant will be able to utilize users' personal information to answer questions and perform actions across various applications, while also having the highest level of privacy protection.

Integrating large AI models with consumer electronics such as PCs and smartphones to create models that can run offline with increasingly powerful inference performance on local devices, while also being able to call upon vast cloud AI computing resources to adapt to users' deeper personal needs, has become the core content of the "AI planning blueprint" for many global consumer electronics companies, including Apple.

Ruoming Pang's departure is one of the most significant personnel losses for Apple's AI team since the launch of the Apple Intelligence super AI project several years ago, highlighting the intensifying competition for AI talent. Meta has offered this top engineer a salary far exceeding that of similar positions within Apple.

Insiders say that Ruoming Pang's departure may trigger a chain reaction of resignations among the core members of the AFM team, with several engineers informing colleagues of their plans to switch jobs to Meta or other companies in the near future. According to previous media reports, Pang's capable assistant Tom Gunter left Apple last month.

The AFM team reports to Daphne Luong, the deputy to AI Senior Vice President John Giannandrea. It is reported that earlier this year, Giannandrea was marginalized internally, and the consumer product-related teams he was responsible for, including Siri, robotics, Core ML, and App Intents, were stripped away This stems from the lackluster response to Apple Intelligence and the continued delays in the new Siri features (including executing commands using user data).

According to insiders, after Pang's departure, AFM will be taken over by Zhifeng Chen. Unlike during Pang's tenure, when most engineers reported directly to him, the new structure will establish multiple managers reporting to Zhifeng Chen, who will then manage the engineers. Team insiders indicate that Chong Wang, Zirui Wang, Chung-Cheng Chiu, and Guoli Yin may serve as the new managers.

Apple's overall AI-related business strategy is currently primarily led by Apple Software Engineering head Craig Federighi and Mike Rockwell, the head of Siri engineering who developed Vision Pro. Jaan Andriya is fully responsible for the AI research department. At the Worldwide Developers Conference in June, Apple's self-developed AI features only appeared in the new call and text translation functionalities.

Other important AI features, such as device screenshot analysis and enhanced image generation, come from partners like OpenAI and Google. Apple also released a new version of Xcode that enables code completion through the Claude and ChatGPT applications