Our approach to alignment research

OpenAI describes its approach to alignment research, focusing on improving AI systems' ability to learn from human feedback and assist humans in evaluating AI, with the goal of building a sufficiently aligned AI system to solve other alignment problems.