Radical Talks

Building Spatial Intelligence: How World Labs is Creating the Next Frontier in AI

Featured speakers: Fei-Fei Li, CEO and Co-founder of World Labs

http://Radical%20Talks%20Episode%20Initial%20Concepts

From our latest Radical Talks episode with World Labs CEO and Co-founder Fei-Fei Li

The AI revolution has transformed language, images, and video generation, yet one fundamental dimension remains largely unconquered: the 3D spatial world. While language models excel at conversation and diffusion models create stunning images, AI still struggles to understand and generate the three-dimensional spaces that humans navigate effortlessly every day.

World Labs, founded by renowned AI researcher Fei-Fei Li, is tackling this frontier by developing spatial intelligence — AI systems that can reason about, generate, and interact with 3D environments. 

This conversation is particularly timely as World Labs commercially launched Marble last week, their generative world model that transforms text prompts, photos, videos, or panoramas into fully editable, downloadable 3D environments.

The Missing Piece in AI

Li’s career has been defined by pursuing audacious problems. Her previous North Star was visual intelligence — for over 15 years, she worked on teaching machines to understand the visual world, culminating in ImageNet, which helped catalyze the deep learning revolution.

Now, her North Star has evolved to spatial intelligence: building AI that can reason, interact, and understand in 3D space. Looking at generative AI’s trajectory, Li sees a clear pattern of breakthroughs in language models, speech, music, multimodal systems, and image and video generation. “If you look at [those breakthroughs], there is a missing piece,” Li notes. “The missing piece is the world.”

This matters because spatial intelligence is fundamental: the ability to navigate and interact in 3D space, whether cooking, playing sports, or dancing, sits at the heart of intelligence. Li explains that spatial intelligence is about “the deep understanding and intelligence about 3D space.”

And it’s not just a technical gap — it’s a massive business opportunity. Industries from design and creativity to robotics, AR, and VR desperately need better 3D tools. The combination of this need and the technological opportunity made starting World Labs irresistible for Li and her co-founders.

Building at the Frontier

World Labs is a frontier lab, working on deep science rather than scaling existing approaches. The challenges are formidable: generating 3D worlds that respect geometry, allow navigation, and enable interaction. “Every day we feel like we’re at the very edge of what has been known, and we have to push that envelope,” Li says.

Recently, World Labs shared a technical blog and opened a viewer called Marble, allowing people to experience what the model can do. The work is guided by a clear philosophy: human-centered AI that empowers rather than replaces.

“I firmly believe creativity is controllability,” Li emphasizes. “Creators have stories and images in their head that are so unique, and what they need are tools that allow them the controllability to create the magic.” This is why 3D is so powerful — it gives creators unprecedented control over cameras, movement, and content. Li invokes Einstein: “Creativity is intelligence having fun.” Spatial intelligence combines creativity, intelligence, productivity, and experience in new ways.

Applications span game development, VFX, filmmaking, design, robotics, VR and more. Li shares an example: co-founder Ben Mildenhall recently created a world where you shoot balls at a Jenga tower. “It’s still pretty rudimentary, but I was playing with it earlier today. I was like, ‘Oh, this is really fun,” shares Li.

Like language models, world models represent a horizontal technology that will enable applications no one has yet imagined. “I’m really proud that I made data cool in AI,” Li says. “Now I think World Labs is going to make 3D cool again.”

The Path Forward

Building spatial intelligence represents one of AI’s remaining frontiers. Just as ImageNet enabled visual recognition and language models transformed NLP, world models could reshape how humans design, create, and interact with virtual and physical environments.

The technical challenges remain formidable, but for Li, that’s precisely the point. She’s drawn to problems hard enough to matter and exciting enough to sustain passionate pursuit. As World Labs continues its journey, the team remains focused on building AI that empowers people, enhances creativity, and expands human capability.

This post is based on insights from Radical Talks, a podcast from Radical Ventures exploring innovation at the forefront of AI. For more cutting-edge AI discussions, subscribe wherever you get your podcasts.