Ilya Sutskever’s Team at OpenAI Built Tools to Control a Superhuman AI

OpenAI has released a research paper, indicating that they believe superhuman AI could be developed within the next ten years. The paper, “Weak-to-Strong Generalization”, outlines a proposed solution to ensure that AI systems much smarter than humans will still follow human rules. OpenAI plans to achieve this by designing small AI models to teach superhuman AI models. The study found that training large AI models with smaller AI models results in a higher level of accuracy in multiple circumstances compared to human training.