Why Video Agent models are next, Ethan He, xAI Grok Imagine Lead

Ethan He asserts that Video Agent models derive their intelligence primarily from large language models, not from training on video data. He argues that the next frontier for interactive, real-time world models lies in advancing LLMs, possibly through Interaction Models. The next Sora-like model will not be a better video model but an improved LLM. He made this claim during a Latent Space session while leading xAI's Grok Imagine development. This perspective shifts focus from video data training to language-based reasoning for video agents.