The podcast discusses the release of Sora 2, OpenAI's new video generation model, highlighting key details often missed in the initial announcements. It explores the existence of two versions, Sora 2 and Sora 2 Pro, and the deliberate, safety-focused rollout strategy. The podcast also compares Sora 2 with Google's VEO 3, emphasizing the importance of training data in model performance. It delves into OpenAI's upcoming social media app, Sora, including its unique cameo feature and the company's promise to discontinue the service if it doesn't benefit users' well-being. The discussion then shifts to Periodic Labs' approach to automating scientific experiments and Claude 4.5 Sonnet's coding capabilities, ending with a reflection on the future implications of AI models that can create indistinguishable realities.
Sign in to continue reading, translating and more.
Continue