So the top two stories are about a model that can generate astonishingly good video from text and a model that has a context window which allows it to process a

So the top two stories are about a model that can generate astonishingly good video from text and a model that has a context window which allows it to process and identify nuanced details in an hour long video. We've fairly quickly moved from a world where AIs would communicate with each other through text to one in which they can do so through video. I'm very curious how something like Sora might end up being used to generate synthetic training data for multimodal models...