The members of OpenAI’s Superalignment team were busy working on the challenge of controlling AI that is more intelligent than humans. The team, led by Ilya Sutskever, is focused on developing ways to steer, regulate, and govern “superintelligent” AI systems. The team’s work is being closely watched, especially in light of recent events at OpenAI.
Sutskever was among those who pushed for the firing of Sam Altman, the former leader of OpenAI, who recently returned. The Superalignment team’s efforts aim to address issues related to the emergence of superhuman AI systems and their impact on humanity. The team is currently exploring governance and control frameworks for future powerful AI systems.
One approach involves using a weaker, less-sophisticated AI model to guide a more advanced model in desirable directions. This analogy helps the team test its superalignment hypotheses and address issues such as algorithmic bias and AI’s tendency toward toxicity. OpenAI is also launching a $10 million grant program to support technical research on superintelligent alignment, with funding from figure Eric Schmidt.
The involvement of Schmidt, who has commercial motivations, raises questions about the availability and use of OpenAI’s research. However, OpenAI has assured that its research and the work of others who receive grants and prizes for superalignment-related projects will be shared publicly. The company’s commitment to making its research publicly available will be closely monitored.