multimodal AI

“I think it's inevitable that in about two or three years, we will have models that are multimodal, large context windows, are capable of reasoning over, say, three hours of a screen recording.”