Nobody Agrees What a 'World Model' Is — Justin Johnson (World Labs) Explains Why
The TWIML AI Podcast with Sam Charrington1 September 2026Watch on YouTube
What you'll learn
- There is no shared definition of 'world model'; the term covers at least three meanings that run together.
- The POMDP framing helps to slot almost any current world model into the roles of planner, simulator, or renderer.
- World Labs deliberately builds two directions: the explicit Gaussian splat path of Marble and the implicit pixels-only path of RTFM.
- With enough data and compute, Johnson believes the implicit 3D route scales up to infinity.
- The real challenge is handling extremely long contexts of hundreds of thousands to millions of tokens.
Topics
Read next
World Labs genereert duizenden gesimuleerde robottaken voor training
World Labs, opgericht door AI-onderzoeker Fei-Fei Li, presenteert een simulatie-engine die robotcontrollers trainen in virtuele omgevingen op basis van reële taken.
AI-labs hebben geen openbare plannen voor insluiting van ontregelde modellen
Een nieuwe studie toont aan dat toonaangevende AI-labs weinig publiek gedocumenteerde plannen hebben voor het insluiten van rogue models.
Sources
What is known about this topic outside the broadcast, and where it says so.
Description from the channel
In this episode, Justin Johnson, co-founder of World Labs, joins us to discuss world models and the emerging field of spatial AI. We explore why many researchers see capabilities beyond language as an important frontier for AI, and what it means to build models that can understand, generate, and simulate the environments around them. Justin explains the different approaches to world modeling, including explicit 3D representations and generative models, and why there is still no established recipe for building these systems. We also discuss World Labs’ Marble system, which can generate navigable 3D worlds from images and other inputs, the challenges of evaluating world models, and the role of simulation, planning, and action. Finally, Justin shares his vision for models that bring these capabilities together, supporting everything from interactive virtual environments to agents and robots that can operate in the physical world. 🗒️ Full show notes including references: https://twimlai.com/go/775. 🔔 Subscribe to our channel for more great content just like this: https://youtube.com/twimlai?sub_confirmation=1 📖 CHAPTERS =============================== 00:00 - Introduction 03:19 - Defining World Models 08:10 - World Models as Theory Builders 11:37 - POMDPs and Agent–World Interaction 15:40 - Training Agents with Behavior Cloning 18:40 - Ground-Truth State and Learned State 23:34 - Explicit 3D vs. Implicit 3D 28:14 - Reconstruction vs. Generative World Modeling 30:38 - Gaussian Splat Anatomy and File Formats 33:52 - Why Gaussian Splats Work with Neural Networks 37:00 - Consistency by Construction and at Scale 40:31 - How Marble Generates 3D Worlds 44:25 - Training Data and Output Representations 47:48 - World Models as Renderers, Planners, and Simulators 51:30 - When Rendering and Simulation Overlap 56:28 - The Path to Unified World Models 59:30 - Architectures, Loss Functions, and Long Contexts 01:02:47 - Where to Learn More 🗣️ CONNECT WITH US! =============================== Subscribe to the TWIML AI Podcast: https://twimlai.com/podcast/twimlai/ Follow us on Twitter: https://twitter.com/twimlai Follow us on LinkedIn: https://www.linkedin.com/company/twimlai/ Join our Slack Community: https://twimlai.com/community/ Subscribe to our newsletter: https://twimlai.com/newsletter/ Want to get in touch? Send us a message: https://twimlai.com/contact/