OpenAI Dissolves Superalignment Team
The dissolution of the team came on the heels of the resignations of its co-leaders, Sutskever and Leike.
OpenAI has dissolved its Superalignment team, originally established to address the potential risks associated with advanced AI systems, a surprising move coinciding with the launch of its latest AI iteration, GPT-4o.[1]
Read More: OpenAI Staff Demand Resignation of Board
The Superalignment team, spearheaded by Ilya Sutskever and Jan Leike since its inception in July 2023, was dedicated to the critical task of mitigating the hazards of AI, including the ominous specter of AI "going rogue."
The dissolution of the team came on the heels of the resignations of its co-leaders, Sutskever and Leike.
But over the past years, safety culture and processes have taken a backseat to shiny products.
— Jan Leike (@janleike) May 17, 2024
Leike, in announcing his departure, cited longstanding disagreements over the company's core priorities, lamenting the Superalignment team's struggles in securing resources for their research endeavors. Their mission, outlined last July, aimed to allocate a significant portion of OpenAI's computing power toward developing a human-level automated alignment researcher over four years.