Taking Charge: OpenAI's New Team Aims to Steer 'Superintelligent' AI Systems

Exciting news from OpenAI! They're forming a new team, led by Ilya Sutskever, to develop ways to steer and control "superintelligent" AI systems.

They predict that AI surpassing human intelligence could arrive within the decade, requiring research into controlling and restricting it. The new Superalignment team, with access to 20% of OpenAI's computing resources, aims to solve the core technical challenges of controlling superintelligent AI over the next four years. Their goal? To build a "human-level automated alignment researcher." 

OpenAI believes that AI systems can make faster progress in alignment research than humans. As AI advances, it can take over more alignment work, conceive better techniques, and collaborate with humans to ensure alignment with human values. 

Collaboration is key. Human researchers will focus on reviewing alignment research done by AI systems, rather than generating it themselves. OpenAI also plans to share their efforts widely and contribute to the alignment and safety of non-OpenAI models. 

 Challenges lie ahead. OpenAI acknowledges limitations and potential biases in using AI for evaluation. However, they believe that tackling superintelligence alignment is a critical machine learning problem and are committed to working on it.