Meta crazily poaches four top Chinese talents from OpenAI
In the past week, eight top researchers have left OpenAI to join Meta, including four core researchers of Chinese descent, all of whom were leaders of core projects at OpenAI, including key models such as o3 and the GPT-4 series. The high-intensity work of 80 hours per week at OpenAI has created opportunities for competitors. The company's Chief Research Officer Mark Chen stated in an internal memo, "It's like someone broke into our house and stole something."
The talent war in Silicon Valley has escalated, with Meta launching a strong "poaching campaign," and OpenAI exclaiming: We've been robbed!
According to media reports, OpenAI is experiencing an unprecedented talent exodus crisis. In the past week, eight top researchers have left to join Meta, including four core Chinese researchers, all of whom were leaders of key OpenAI projects, including o3, the GPT-4 series, and other critical models.
OpenAI CEO Sam Altman previously revealed in a podcast that Meta offered some researchers signing bonuses of up to $100 million. However, Meta executives have refuted this figure internally.
OpenAI is facing a challenge in retaining talent. OpenAI's Chief Research Officer Mark Chen stated in an internal memo sent to employees on Saturday that the company is "recalibrating compensation" and promised to take "creative ways to recognize and reward top talent."
According to Wired, Chen expressed extreme shock and dissatisfaction in the memo, writing: "I have this gut feeling right now, like someone has broken into our home and stolen something."
OpenAI's Work Pressure is Too High, Meta Seizes the Opportunity to "Poach"
According to media reports, Meta is significantly ramping up its recruitment efforts, particularly focusing on talent from OpenAI and Google. Zuckerberg has been taking a particularly aggressive recruitment strategy, even personally reaching out to potential recruits.
Reports indicate that OpenAI employees face high-intensity work pressure of 80 hours a week, which has created opportunities for competitors to recruit. Chen also acknowledged in the memo that the company has been too focused on regular product releases and short-term comparisons with competitors, and now needs to refocus on achieving the "primary task" of general artificial intelligence.
To stabilize morale, Chen promised to confront this talent war head-on with CEO Sam Altman and other leaders. He explicitly stated in the memo that the leadership "is not standing by," but is taking action with an unprecedented proactive attitude, with core measures including reassessing and adjusting the compensation system and designing more creative incentive programs. However, Chen also emphasized that he has "very high personal fairness standards" and will not compromise fairness for other employees to retain certain individuals.
The memo also included information from seven other research leaders, urging employees to remain calm and to communicate with management when receiving offers from Meta.
The talent exodus has forced OpenAI to reflect on its internal management and strategic direction. To alleviate employee pressure, OpenAI has decided to arrange a week of collective leave for employees to "recharge." However, one leader warned in the memo that Meta is aware that OpenAI will be on collective leave this week to "recharge," and may take this opportunity to apply pressure and make "absurd, time-limited explosive offers."
Four Chinese Generals, a Major Loss for OpenAI
The four Chinese researchers poached by Meta are all recognized technical backbones within OpenAI, including Jiahui Yu, the leader of the o3, o4-mini, and GPT-4.1 projects; Hongyu Ren, the creator of the o3-mini and o1-mini models; Shuchao Bi, the head of the multimodal model post-training team; and Shengjia Zhao, a key contributor to the GPT-4 and o1 projects.
Jiahui Yu is the Director of Perception Research at OpenAI, having graduated with a bachelor's degree from the University of Science and Technology of China and a Ph.D. from the University of Illinois at Urbana-Champaign. He joined Google Brain in 2019, moved to Google DeepMind in 2020, and joined OpenAI in October 2023. During his time at OpenAI, Yu led the o3, o4-mini, and GPT-4.1 projects, served as a research advisor for GPT-4o image generation, and was responsible for the development of the "Thinking with Images" image reasoning project.
Hongyu Ren is the creator of the o3-mini and o1-mini models, having graduated with a bachelor's degree from Peking University and a Ph.D. from Stanford University. He joined OpenAI in July 2023. He is also a core contributor to the o1 project and the head of GPT-4o mini, primarily responsible for the post-training team, focusing on training models to think faster, deeper, and more sharply.
Shuchao Bi serves as the head of the multimodal model post-training team at OpenAI, having graduated with a bachelor's degree from Zhejiang University and a Ph.D. from the University of California, Berkeley. He previously served as an engineering director at Google and was a co-founder of YouTube Shorts. His research at OpenAI covers cutting-edge technologies such as new paradigms for post-training, multimodal reasoning, and the development of multimodal agents.
Shengjia Zhao is a key contributor to the GPT-4 and o1 projects, having graduated with a bachelor's degree from Tsinghua University and a Ph.D. from Stanford. He joined OpenAI in June 2022, focusing on the development of ChatGPT and the GPT-next projects.
