文章来源：钛媒体
TMTPOST -- Apple has lost four high-profile artificial intelligence (AI) researchers in a development that highlights an ongoing trend: top talent in the AI sector is highly mobile, and high salaries alone aren』t the only reason engineers change companies.
According to a report by Bloomberg journalist Mark Gurman, Apple』s recent departures include Jian Zhang, the company』s Chief AI Researcher for Robotics, and three key members of its Foundation Models team: Nan Du, Zhao Meng, and John Peebles.
The departures reveal two critical insights. First, the talent loss is highly concentrated, with three researchers from the foundational model group. Second, the exodus underscores the prominence of Chinese talent in the field—three of the four departing employees are Chinese.
Despite comparisons to Meta』s high-profile poaching campaigns, the talent migration wasn』t driven solely by Meta. Of the four, only Jian Zhang joined Meta, while Nan Du and John Peebles moved to OpenAI, and Zhao Meng joined Anthropic.
Jian Zhang』s departure is particularly notable. Zhang joined Apple in 2005, serving a decade as the Head of Robotics Research within Apple』s AI and Machine Learning division. Unlike Tesla』s humanoid robot initiatives, Apple』s robotics research aims to underpin future product lines, from desktop robots with screens to robotic arms for retail stores and manufacturing applications.
Zhang, a Zhejiang University alumnus with a PhD from Purdue University, has a strong academic track record in robotics and biomimetic flapping-wing micro air vehicles. His work includes system integration, geometric flight control, and reinforcement learning applications in robotics. Notably, his paper on 「Uncertainty-Weighted Actor-Critic Algorithms for Offline Reinforcement Learning」 has been cited more than 240 times, addressing complex robotic control challenges without real-time environmental interaction.
Bloomberg reported that after leaving Apple, Zhang joined Meta』s new Robotics Studio. Meta』s division aims to develop humanoid robot hardware and software for household scenarios, as well as provide AI, sensors, and software for third-party developers. Although Meta froze hiring for its Superintelligence Lab earlier this year, the Robotics Studio is separate from that unit, leaving questions about the precise incentives behind Zhang』s move.
Regardless, Apple now faces the challenge of losing a decade-long veteran whose work is central to its future robotics ambitions.
Even more concerning is the departure of three key members of Apple』s Foundation Models team. Nan Du, John Peebles, and Zhao Meng moved to different AI powerhouses, signaling a broader trend in the industry.
John Peebles and Nan Du joined OpenAI. Peebles specializes in generative AI and large language models (LLMs), with expertise in deep learning and privacy-preserving AI. He previously contributed to Apple』s foundational model initiatives, including the deep learning training system AXLearn. Peebles』 familial ties to OpenAI—his brother works on the Sora team—likely influenced his move.
Nan Du, who spent over seven years at Google, has been instrumental in projects like GLAM, a trillion-parameter Mixture of Experts model, and PaLM 2, Google』s second-generation Pathways Language Model. Du』s research focuses on computational efficiency, model performance, and generative search technology. At Apple, he contributed to the development of large-scale, efficient foundational models.
Zhao Meng, meanwhile, joined Anthropic. His research spans multimodal AI and generative models, with high-impact work on image-text fusion and knowledge transfer in natural language processing (NLP). Zhao has published papers cited over 770 times, demonstrating influence in zero-shot learning and pre-trained language models. His move to Anthropic coincides with the company』s recent $1.3 billion Series F funding, valuing the AI startup at $183 billion.
The departures illustrate that high compensation isn』t the sole driver for AI talent movement. While Meta is known for offering large multi-year packages—most famously a $200 million offer to poach Apple』s Pang Ruoming—employees often leave for reasons beyond immediate salary.
OpenAI CEO Sam Altman has famously said, 「Missionaries will beat mercenaries,」 emphasizing that researchers are motivated by mission alignment, research autonomy, and work environment. Wired recently reported that two researchers who briefly joined Meta returned to OpenAI within 30 days, suggesting that corporate culture and research direction play decisive roles.
Elon Musk also highlighted this dynamic, noting that xAI recruited senior engineers from Meta despite modest initial pay, demonstrating that mission and project focus can outweigh financial incentives.
The financial stakes in AI are enormous. According to Business Insider, top OpenAI researchers can earn over $10 million annually, while DeepMind offers compensation packages up to $20 million, with special equity grants. These figures underscore that while base salary is high across the board, the relative appeal of a company often hinges on research freedom, vision, and team environment.
Apple』s AI exodus is significant because it involves both core foundational model researchers and robotics experts. Unlike a one-off talent poaching incident, this represents a structural challenge. Apple』s AI ambitions—including robotics integration and large language model development—may face delays as a result.
Furthermore, the concentration of Chinese talent leaving Apple could have broader implications for the company』s research pipeline and diversity strategy. The departures to OpenAI, Anthropic, and Meta highlight the competitive landscape of AI, where even industry leaders must continuously innovate to retain top talent.
The trend suggests that the AI talent war is broader than Meta』s headline-grabbing salary offers. Researchers evaluate company missions, team culture, and project impact, and they are willing to move quickly if another organization better aligns with their professional goals.
For Apple, the challenge now is not just to counteract salary-driven poaching but to foster a research environment that retains top-tier AI talent. This includes clear project roadmaps, meaningful autonomy, and recognition of researchers』 contributions.