World models

Google DeepMind, Stanford professor Fei-Fei Li’s World Labs, and OpenAI are advancing research into AI world models to overcome limitations of large language models (LLMs). These models aim to help AI systems represent the external world and predict action consequences, unlike current LLMs which fail with unexpected variables like detours in navigation. Yann LeCun also left Meta to form a world-model-focused startup, indicating growing industry interest. World models are considered essential for future robotics, with applications ranging from deep-sea exploration to assisting healthcare providers. Google DeepMind and World Labs are currently building models that generate interactive, 3D virtual environments from text, images, and video prompts.