In a move that has raised eyebrows regarding OpenAI’s commitment to mitigating risks from advanced artificial intelligence systems, the company has reportedly dissolved its dedicated “superalignment” team. This decision comes in the wake of recent high-profile departures, including those of the team’s leaders.
According to a report from Bloomberg, OpenAI has opted to integrate members of the superalignment team, which was formed less than a year ago, into the company’s overall research efforts rather than maintaining it as a separate entity. The rationale provided is to help further OpenAI’s AI safety goals while continuing to develop cutting-edge technologies.
The superalignment team was spearheaded by OpenAI’s co-founder and former chief scientist Ilya Sutskever and seasoned researcher Jan Leike. However, their recent resignations, along with those of other team members, have cast doubt on OpenAI’s approach to balancing the pace of AI development with ensuring its safe and responsible deployment.
Sutskever’s departure is said to have stemmed from disagreements with CEO Sam Altman over the urgency of advancing AI capabilities. Leike followed suit shortly after, citing his own misalignments with the company’s direction and challenges securing resources for the superalignment team’s work.
With the core leadership gone and resources dwindling, the future of OpenAI’s dedicated AI alignment efforts appeared uncertain – prompting the decision to dissolve the specialized team altogether.
In its place, OpenAI has appointed John Schulman, a co-founder focused on large language models, to lead the organization’s AI alignment research integrated across teams. The company maintains it has other employees working on AI safety across various groups, including a “preparedness team” tasked with analyzing and mitigating catastrophic risks.
Nonetheless, the changes have fueled concerns about whether OpenAI is truly prioritizing the potential existential risks of advanced AI development. In a recent podcast interview, Altman advocated for an international AI regulatory body to oversee the technology’s evolution and prevent “significant global harm.”
As one of the world’s leading AI research labs, OpenAI’s actions regarding pivotal safety work like the superalignment team will undoubtedly face heightened scrutiny. The company must now prove its integrated approach can uphold its core mission of ensuring artificial general intelligence remains safe and beneficial.
