Reinforcement learning has seen significant success in recent years, but its low sample efficiency limits its application in the real world. World models, as a type of environment generation model, offer hope for addressing this issue. They can serve as simulated environments to train reinforcement learning agents with higher sample efficiency. Currently, most world models simulate environment dynamics through discrete latent variable sequences. However, this method of compressing into a compact discrete representation may overlook visual details that are critical for reinforcement learning. Meanwhile, diffusion models have become prominent in the field of image generation.

Recently, Jige Technology proposed a new framework called DriveDreamer4D, aimed at leveraging prior knowledge from world models to improve the reconstruction quality of 4D driving scenes. Traditional 4D scene reconstruction methods primarily rely on two main approaches: NeRF and 3DGS. NeRF is like a super painter, capable of rendering a collection of photos into a 3D model using neural networks. On the other hand, 3DGS simulates various objects in the scene using a set of three-dimensional Gaussian functions. However, both methods have their limitations.

Beijing Zhiyuan Artificial Intelligence Research Institute announces the launch of the native multimodal world model Emu3. This model is based on next-token prediction technology and does not rely on diffusion models or combinatorial methods to achieve understanding and generation across text, image, and video modalities. Emu3 surpasses existing well-known open-source models such as SDXL, LLaVA, and OpenSora in tasks like image generation, video generation, and visual language understanding, showcasing exceptional performance.

Robot startup 1X Technologies has developed a new generative model that enhances the efficiency of training robotic systems in simulations. The company announced this model in a new blog post, addressing one of the significant challenges in robotics, which is learning the 'World Model (1X World Model)' that predicts how the world responds to the actions of robots. When robots start from the same sequence of initial images, this world model can envision multiple possibilities based on different action proposals.

Norwegian startup 1X Technologies recently announced significant advancements in developing AI-based world models for robots. Simply put, these models act as virtual testing grounds for robots, allowing them to test and improve in various scenarios without the need for real-world testing. 1X believes this is key to solving the 'robotics challenge' — how to reliably evaluate robots trained for multiple tasks in ever-changing environments.

Title: Artificial Intelligence "Pioneer" Fei-Fei Li Founders World Labs, Raises $230 Million! Recently, Stanford University professor and "AI Pioneer" Fei-Fei Li's new startup World Labs successfully raised $230 million, with investors including renowned venture capital firms Andreessen Horowitz, NEA, and Radical Ventures. This funding values World Labs at over $1 billion.

Google has released the base world model Genie, which features 11 billion parameters and can generate interactive virtual environments. Genie trains without action labels, autonomously identifying different action features and patterns in videos. Genie can learn fine-grained control and simulate a variety of potential actions. It generates video in a cartoon style and can simulate robotic movements and deformable objects, leaning more towards animated graphics compared to Sora.