OpenAI's new text-to-video AI model, Sora, can generate realistic scenes, but details about training dodged.

From Wall Street Journal:

OpenAI’s new text-to-video AI model, Sora, can generate realistic scenes. The company’s CTO, Mira Murati, explains how it works in an interview with WSJ’s Joanna Stern. However, questions about how the model was trained were dodged. Photo illustration by Preston Jessee for The Wall Street Journal.



Read more at Wall Street Journal: OpenAI Made Me Crazy Videos—Then the CTO Answered (Most of) My Questions