Fei-Fei Li’s World Labs has introduced Atlas, a new multimodal “world model” built to create, rebuild, and simulate 3D environments using text, images, and video.
The announcement marks a major step for World Labs, the artificial intelligence startup founded by Stanford professor and renowned AI researcher Fei-Fei Li. Atlas is designed to help machines better understand the physical world, moving beyond flat images and text-based outputs toward interactive three-dimensional spaces.
Unlike traditional AI tools that generate only pictures or written responses, Atlas focuses on spatial intelligence. That means it can interpret visual information, understand the structure of a scene, and produce 3D environments that can potentially be explored, edited, or simulated. Users may be able to describe a place in words, provide reference images, or use video input to help the system build a digital world.
This type of technology could have a major impact on gaming, robotics, virtual reality, architecture, film production, simulation training, and AI research. A system that can generate realistic 3D spaces from simple prompts could reduce the time and cost needed to build digital environments. It could also help robots and autonomous systems better understand how objects, rooms, and outdoor spaces relate to one another.
Fei-Fei Li has long emphasized the importance of visual intelligence in artificial intelligence. With Atlas, World Labs appears to be pushing that idea further by developing AI that does not just recognize objects, but understands environments in a more human-like way. The goal is to give AI systems a deeper grasp of the physical world, including depth, layout, movement, and interaction.
Atlas arrives at a time when interest in world models is growing quickly across the AI industry. These systems are seen as an important foundation for the next generation of artificial intelligence because they allow machines to simulate real-world scenarios rather than simply predict text or generate images. By learning how the world works, AI models could become more useful for planning, reasoning, and decision-making.
For creators, Atlas could open the door to faster 3D content generation. A designer might turn a written idea into a usable environment, while a filmmaker could quickly visualize a scene before production. Game developers could use similar tools to prototype levels or create immersive worlds with less manual modeling. In education and training, simulated environments could help students and professionals practice complex tasks in realistic digital settings.
The technology may also be especially important for robotics. Robots need to understand three-dimensional space in order to move safely, handle objects, and operate in homes, factories, hospitals, and public areas. A powerful world model could help train robots in simulation before they are deployed in the real world.
While Atlas is still part of an emerging field, its introduction highlights a broader shift in AI development. The focus is moving from systems that simply generate content to systems that can understand and simulate reality. If successful, world models like Atlas could become a key building block for more advanced AI applications.
World Labs’ debut of Atlas positions the company as one of the most closely watched players in spatial AI. With Fei-Fei Li’s background in computer vision and AI research, the startup is aiming to shape how machines perceive and interact with the world around them.
Atlas represents more than another AI generation tool. It points toward a future where artificial intelligence can build dynamic 3D worlds, reason about physical spaces, and support new forms of creativity, simulation, and automation.
