Fei-Fei Li: Spatial Intelligence Solves Robotics Data Scarcity

Generated: 2026-08-02 · API: Gemini 2.5 Flash · Modes: Summary

Fei-Fei Li: Spatial Intelligence Solves Robotics Data Scarcity

Clip title: Fei-Fei Li is Solving the Hardest Problem in Robotics | World Labs with a16z Author / channel: a16z URL: https://www.youtube.com/watch?v=-tabaM5l3s0

Summary

This video features an interview with Feifei Li, CEO of World AI, and Yunzhu Li, co-founder of Scenix, discussing World AI’s acquisition of Scenix and their combined vision for the future of artificial intelligence. The main topic revolves around building the “next frontier of AI” which they term “Spatial Intelligence.” This involves creating AI that can generate, understand, reason with, and interact with spaces, whether physical or virtual, and a key means to achieve this is through building “large world models.” The acquisition of Scenix is presented as a strategic move to integrate robotics into this vision, enabling AI to act within these spaces.

Yunzhu Li explains that Scenix is developing a “real-to-sim-to-real” pipeline. This pipeline maps real-world environments into highly aligned digital twins, where events in the digital realm accurately reflect those in the physical world. This approach is designed to overcome significant bottlenecks in robotics development, particularly around training and evaluation, by generating vast amounts of high-quality, consistent synthetic data. Consistency over space, time, viewpoints, and types of interactions is crucial for effective robot learning within these digital worlds. World AI’s existing “Marble” generative model, which creates geometrically consistent 3D worlds from prompts, is highlighted as a foundational technology that complements Scenix’s reconstruction capabilities.

The synergistic integration of World AI’s generative models with Scenix’s simulation expertise addresses the profound challenge of data scarcity in robotics. Simulation, unlike real-world data collection, allows for “counterfactual reasoning” – playing out events that haven’t happened or are difficult to observe in reality. This provides two major benefits: reliability, through systematic randomization and coverage of diverse scenarios for robust robot behavior; and efficiency, by drastically speeding up training and evaluation processes compared to slow, costly, and potentially dangerous real-world robot trials. The goal is to train robots for “semi-structured” environments like warehouses and factories, before moving towards fully unstructured domestic settings.

Ultimately, the collaboration aims to build omni-models for robotics, which can process multi-modal inputs (text, image, depth) and produce multi-modal outputs (actions, world states). This creates a scalable, flexible, and robust infrastructure for developing robotic intelligence. The combined team emphasizes a pragmatic, systems-level approach, acknowledging the complexity of creating truly capable robots that operate reliably in the real world, a challenge that requires integrating insights from physics, learning, and extensive data, both real and simulated. They express “measured optimism” about achieving human-level robotic efficiency and capability, recognizing it will be an iterative, long-term endeavor.

Video Description & Links

Description

Last week, World Labs announced its acquisition of SceniX, bringing together two teams working on one of AI’s biggest unsolved problems: how to give machines a true understanding of the physical world.

Martin Casado sits down with Fei-Fei Li, co-founder and CEO of World Labs, creator of ImageNet, and pioneer of spatial intelligence, alongside Yunzhu Li, co-founder of SceniX and assistant professor at Columbia University. They discuss why World Labs acquired SceniX, how simulation can unlock the next generation of robotics, and why training robots may require a fundamentally different approach than training language models.

The conversation explores real-to-sim-to-real pipelines, world models, robotics foundation models, evaluation, synthetic data, and why the future of AI depends not just on understanding language—but on understanding and interacting with the physical world.

Timestamps: 00:00 - Intro 01:08 - World Labs & SceniX 06:04 - Marble & the Data Bottleneck in Robotics 07:13 - How the Two Teams Come Together 10:55 - Building a Foundation Model for Robotics 12:35 - Video Models vs Real-to-Sim-to-Real 19:38 - Why Simulation is Essential for Robot Learning 23:01 - Training, Evaluation & Real Customer Use Cases 29:17 - Humanoids, Semi-Structured Environments & the Grand Challenge 36:56 - Integration Plans & What Success Looks Like in Two Years

Resources: Follow Fei-Fei Li on X: https://x.com/drfeifei Follow Yunzhu Li on X: https://x.com/YunzhuLiYZ Follow Martin Casado on X: https://x.com/martin_casado

Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends!

Find a16z on X: https://twitter.com/a16z

Find a16z on LinkedIn: https://www.linkedin.com/company/a16z

Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX

Listen to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711

Follow our host: https://x.com/eriktorenberg

Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures.

NemoClaw Knowledge Wiki

Explorer

Fei-Fei Li: Spatial Intelligence Solves Robotics Data Scarcity

Fei-Fei Li: Spatial Intelligence Solves Robotics Data Scarcity