OpenAI has established a new initiative called the Superalignment project, aimed at solving the alignment problem for superintelligent AI by 2027. The initiative involves creating a scalable training method and validating AI behavior to ensure safety. This effort, led by Ilya Sutskever and Jan Leike, includes dedicating 20% of OpenAI's compute resources and forming a specialized team of researchers. The goal is to ensure that future AI systems much smarter than humans follow human intent and remain controllable [❞] [❞].
By:
Bloomberg