Robotics · 2026-11-08
World Models: How AI Learns to Understand the Physical World, and Why Robots and Cars Care
World models are AI systems that learn to predict how the real world changes, which could make robots and self-driving cars smarter.
What a world model is in simple terms
Most well-known AI systems, such as ChatGPT, work with text: they predict the next word in a sentence. World models work differently. They take a visual picture of the world and predict what the next state will be. In other words, such a model tries to understand what happens next when something moves, falls, or changes.
The key difference is that a world model has to account for physics, cause and effect, and how people and other objects react to our actions. That is why it is called a world model rather than just a language model.
Why robots and cars need virtual practice
World models can serve as virtual training grounds. According to Brian McClendon, chief technology officer of Niantic Spatial, someone could photograph the place where a robot is meant to work, generate a 3D model, and let the software run "millions of simulations." The robot would learn the space's parameters without physically moving through it.
Driverless cars also benefit, because without a human driver, the software has to react in split seconds. Wayve builds world models for cars. It uses, in part, dashcam footage to recreate hazardous roads and unsafe conditions that would be impractical or dangerous to test in real life.
Where it is already being used
Beyond robots and cars, world models interest video game developers, since physics and 3D rendering are core to games. The Chinese tech giant Tencent is investing in its own world models.
There is also a less visible effect: training these systems requires real recordings of everyday life. The article's author recorded himself folding laundry and washing dishes on his phone and sold the videos through an app for a modest payment. He suggests that this kind of gig work may grow as more companies build world models.
Useful from the source
- Niantic Spatial — A company built on Niantic's work that is developing world models, including ones based on real-world location data.
- Odyssey — An AI lab developing world models that generate realistic 3D environments.
- Wayve — A startup building world models for autonomous cars.
Sources: 🔗 WIRED — AI