EP 106 : Minute-Long AI Videos, New Amazon Voice & Video Models, and OpenAI Startup Moves

Discover the latest advancements in artificial intelligence with today's AI Brief. We delve into groundbreaking research from NVIDIA and Stanford that enables the generation of minute-long, coherent AI cartoon videos using a new technique called "Test-Time Training". This innovation significantly improves consistency and storytelling capabilities in AI-generated video, outperforming existing methods. Learn how this new approach, demonstrated with Tom and Jerry cartoons, modifies existing video models to handle much longer sequences.We also cover Amazon's latest AI developments, including the launch of Nova Sonic, a new voice model for human-like interactions that boasts lower latency and error rates than comparable OpenAI models. Additionally, Amazon has upgraded its Nova Reels 1.1 video model, extending video generation to 2 minutes with improved quality. Both models are now available through Amazon Bedrock.Stay informed about Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, which has recently added ex-OpenAI CRO Bob McGrew and GPT architect Alec Radford to its advisory team, bringing the number of OpenAI alumni on board to nearly half. Finally, we touch upon other significant AI news, including Gemini's enhanced reasoning capabilities for analyzing real-world data, and highlight some of the trending AI tools available today, such as Gemini Live and Runway Gen-4 Turbo. Tune in to stay ahead in the world of AI!