Recently, OpenAI's Vice President of Chinese origin, Li Weng, announced her departure to be effective on November 15th. Weng graduated from Peking University and was the head of OpenAI's security systems team, also being one of the key contributors to ChatGPT.
She joined OpenAI in 2018 and has been involved in several significant projects, including GPT-4, particularly in areas such as model safety, pre-training, and reinforcement learning.
Throughout her career, Weng proposed the famous formula "Agent = Large Model + Memory + Active Planning + Tool Use," providing crucial theoretical support for AI research and applications. She delivered a keynote speech on "AI Safety and Cultivation" at the 2024 Bilibili Super Science Night event, which garnered widespread attention and discussion.
In her farewell letter, Weng expressed her gratitude towards OpenAI and reflected on her growth journey within the company. She mentioned her initial curiosity and enthusiasm upon joining, participating in projects ranging from deep reinforcement learning algorithms to robotic challenges, and successfully teaching a robotic hand to solve a Rubik's Cube through the team's collaborative efforts. She also noted that while exploring the application of AI models in the real world, she established the first applied research team, laying the groundwork for subsequent safety initiatives.
Under her leadership, OpenAI's security systems team has grown significantly, now comprising over 80 outstanding scientists and engineers. Weng takes pride in the team's achievements in AI safety, particularly the recently released o1 preview model, which demonstrates exceptional safety and practicality. Her stringent requirements for safety testing and red team testing ensured the robustness and transparency of the models.
Although leaving OpenAI, Weng plans to continue updating her blog and looks forward to dedicating more time to programming and writing. Her departure has been met with well-wishes from many colleagues, marking a new beginning in her career and signaling a new appointment for the crucial role of OpenAI's head of security.
Key Points:
🌟 Li Weng announces her departure from OpenAI on November 15th, having served as the head of the security systems team.
📈 During her tenure, she participated in multiple significant projects, including GPT-4, and proposed the renowned "Agent" formula.
💻 Post-departure, Weng plans to continue updating her blog and looks forward to spending more time on programming and writing.