In Short:
Tech experts claim the AI hype is fading, but Fei-Fei Li, known as the “godmother of AI,” disagrees. She co-founded World Labs to advance spatial intelligence, aiming to create immersive worlds from AI. The startup has raised $230 million, with a billion-dollar valuation. Li’s vision could lead to significant advancements in robotics, augmented reality, and autonomous technology.
AI Innovator Fei-Fei Li on New Ventures and Spatial Intelligence
Despite assertions from market-focused technology analysts that the artificial intelligence (AI) bubble has burst and a downturn is imminent, Fei-Fei Li, renowned as the “godmother of AI,” remains optimistic. Currently on a part-time leave from Stanford University, Li is co-founding a startup known as World Labs. While existing generative AI predominantly deals with language, Li envisions a future where systems can construct entire worlds, capturing the complexities of our physical reality with precision and detail. Despite skepticism surrounding AI advancements, World Labs is swiftly moving forward, having secured approximately $230 million in funding, with investors reportedly valuing the startup at around a billion dollars.
Creating New Paradigms in AI with ImageNet
About ten years ago, Li played a pivotal role in transforming AI by developing ImageNet, a specialized database of digital images that significantly enhanced the intelligence of neural networks. She believes that contemporary deep-learning models require a similar foundational boost to enable the creation of authentic worlds, whether replicating real-life environments or fabricating entirely new ones. The creative potential for future storytellers may shift, with authors potentially drafting their narratives as prompts for immersive experiences. As Li articulates, “The physical world for computers is seen through cameras, and the computer brain behind the cameras. Turning that vision into reasoning, generation, and eventual interaction involves understanding the physical structure and dynamics of our world. This technology is termed spatial intelligence.” World Labs identifies itself as a spatial intelligence company, with its success poised to either redefine the term or render it obsolete.
Exploring the Future of Spatial Intelligence
Li has devoted years to researching spatial intelligence. While the mainstream focus was on developments like ChatGPT, she, alongside former student Justin Johnson, was passionately discussing the next phase of AI evolution. Johnson, now an assistant professor at The University of Michigan, explains, “The next decade will be about generating new content that takes computer vision, deep learning, and AI out of the internet realm and embeds them in space and time.”
Partnership with Industry Leaders
The motivation to launch a startup crystallized for Li in early 2023 after a dinner with Martin Casado, a key figure in virtual networking and partner at Andreessen Horowitz. Recognized for its enthusiastic investment in AI, Casado equates the current trajectory of AI to the evolution of computer gaming—from text-based experiences to stunning 3D graphics, with spatial intelligence being the driving force behind this transformation. Casado envisions a future where one can immerse themselves in their favorite fictional narratives in a dynamic, real-time environment. He notes, “The first step to making that happen is progressing from large language models to large world models.”
Assembling a Talented Team for World Labs
Li began to assemble a talented team, including Johnson as a co-founder. Casado also recommended two additional members: Christoph Lassner, an innovator with a background at Amazon, Meta’s Reality Labs, and Epic Games, and the creator of Pulsar, a rendering technique leading to improvements in scene synthesis. Lastly, Casado suggested Ben Mildenhall, known for developing NeRF—neural radiance fields that convert 2D images into compelling 3D visuals. Mildenhall recently left his role as a senior research scientist at Google to join Li’s team.
Future Aspirations of World Labs
A primary objective for World Labs is to integrate advanced world models into robotics, although this will be a longer-term goal. The initial phase focuses on constructing a robust model that comprehensively understands three-dimensionality, physicality, and the intricacies of space and time. Subsequent phases will explore augmented reality, eventually leading to advancements in robotics. If successful, these large world models could revolutionize sectors such as autonomous vehicles, automated manufacturing, and possibly even humanoid robotics.
