OpenAI co-founder Ilya Sutskever and co-lead Jan Leike left the company due to the (alleged) lack of prioritization of security; understand
OpenAI ’s “Superalignment” team , responsible for developing strategies to manage “superintelligent” AI systems , faced significant difficulties. Initially promised 20% of the company’s computing resources, the group had many requests denied. And prevented her from fulfilling her role.
A. Difficulties and disbandments in an important OpenAI team
- OpenAI’s “Superalignment” team, aimed at managing “superintelligent” AI systems, faced significant difficulties, including frequent denials of requests for promised computing resources, which prevented it from fulfilling its role;
- Several team members, including co-lead Jan Leike, a former DeepMind researcher, and Ilya Sutskever, co-founder of OpenAI, left the company due to frustrations with the lack of priority given to security and preparation for future generations of AI;
- The “Superalignment” team contributed significantly to the security of AI models and published important research, but faced difficulties obtaining necessary investments as product launches became a priority. Sutskever and Leike’s departure resulted in John Schulman taking over the team’s projects;
- OpenAI decided not to maintain a team dedicated to AI security, instead integrating these researchers across multiple divisions. This internal rearrangement raises doubts about OpenAI’s commitment to safety in the development of “superintelligent” AI, a concern given the dangerous potential of this technology.
This scenario led to the resignation of several members, including co-leader Jan Leike, a former DeepMind researcher. During his time at OpenAI, Leike played a crucial role in the development of projects such as ChatGPT, GPT-4 and InstructGPT.
