Outburst! OpenAI is facing personnel turmoil again! The head of security has resigned, and the "super-smart alignment team" has been disbanded

National Business Daily

2024-05-18 09:07Published on the official account of Sichuan Daily Economic News

Edited by: Du Yu

In the early morning of May 18, Jan Leike, head of security at OpenAI and head of the "Super Intelligence Alignment Team", announced on social platforms that he was leaving OpenAI. This is the resignation of another executive after the resignation of OpenAI co-founder and chief scientist Ilya Sutskever on Wednesday.

Jan Leike publicly revealed part of the reason for his departure on X. His subtext is that he and OpenAI's top brass have a long-standing disagreement over the company's core priorities, and eventually the differences have become irreconcilable. The team encountered significant hurdles in advancing their research projects and securing computing resources, and the lack of resources could seriously impact the progress and quality of the research.

This Friday (May 17 local time), OpenAI confirmed that the "Super Intelligence Alignment Team" co-led by Ilya and his cronies Jan Leike has been disbanded. Thereafter, the team's research work will be integrated into other OpenAI research groups. At this point, the Gong Dou incident, which once almost led to the collapse of OpenAI, has finally settled.

Musk commented on the news that OpenAI's long-term AI risk team has been disbanded: "This shows that safety is not OpenAI's top priority. ”

OpenAI CEO Sam Altman thanked and affirmed Jan Leike's contributions and leadership at OpenAI. He also said that a longer article will be published in the next few days to explain OpenAI's plans and actions for product security.

Prior to joining OpenAI, Jan Leike worked at Google DeepMind, where he was responsible for prototyping reinforcement learning from human feedback. During his time at OpenAI, he was involved in the development of InstructGPT, ChatGPT, and the secure alignment of GPT-4.

Regarding the resignation of Ilya and Jan, some people said that these two security leaders have resigned, and who will be responsible for OpenAI's product security in the future?

There is also a netizen who broke the news that it is okay to leave OpenAI, but you must sign a "separation agreement", If you do not sign a lifelong non-defamation commitment, you will lose all the OpenAI shares you have acquired. I don't know if Jan Leike's complaints will affect his equity.

In July last year, OpenAI announced the establishment of a new research team "Super Alignment", which will use AI to supervise AI to solve the problem of super-intelligent "AI alignment". "AI alignment" means that the goals of AI systems are aligned with human values and interests. One of the founders of the company, Ilya Sutskovo, was appointed co-leader of the team.

And now the dissolution of the AI risk team is yet another sign of recent turmoil within the company, and has once again raised questions about how the company balances speed and security when developing AI products. OpenAI's charter states that artificial general intelligence (AGI), that is, technology that can match or surpass humans, must be developed safely.

National Business Daily comprehensive public information

View original image 37K

• 

• 

• 

• 

•