“OpenAI believes the next major breakthrough for video models will be the ability to simulate long-duration processes, enabling applications in scientific research for fields like biology and physics.”