Autoencoders

World Models & General Intuition: Khosla's largest bet since LLMs & OpenAI

Latent Space

Used in video models, allowing world models to predict at smaller resolutions which can then be enriched.