World Labs has emerged from stealth with $230 million to unlock spatial reasoning for AI

Leading artificial intelligence scientist Fei-Fei Li, best known for developing the computer vision dataset ImageNet, founded AI startup World Labs with computer vision researchers Justin Johnson, Christoph Lassner, and Ben Mildenhall. World Labs recently emerged from stealth with $230 million in total funding and already boasts a valuation surpassing $1 billion. Andreessen Horowitz, NEA, and Radical Ventures were the main contributors of this initial funding. Additional investors include Marc Benioff, Ashton Kutcher, Adobe Ventures, AMD Ventures, Databricks Ventures, and NVentures, NVIDIA's venture capital arm.

World Labs will tackle the problem of how to enable AI to reason about the physical world, which requires models to go beyond text and image processing to include broader, more abstract types of reasoning, involving understanding how physics works, and concepts such as time, space, and cause-and-effect. World Labs' research will translate into large world models (LWM), intended to be the spatial reasoning counterparts of large language models. Just as the latter can process and analyze text and images giving AI systems something akin to 2D-reasoning capabilities, LWMs will enrich AI systems by endowing them with 3D-reasoning.

According to Andreessen Horowitz, the startup is working on a foundation model for 3D interactive world generation, which the company takes as the basis for spatial reasoning. A clear advantage of World Labs' first model is it has obvious short-term applications, including design, gaming, visual effects, and eventually, robotics.