Waymo Catches World Model Fever, and the Only Prescription Is More World Models

Waymo vehicles have reportedly driven more than 200 million miles autonomously on public roads. But it has not yet encountered a tornado or an elephant, and it is likely that if it did, its reaction would be poor. To try to help with once-in-a-billion-mile scenarios, Waymo announced Friday that it’s introducing the Waymo World Model, a generative AI model that it will use to run through nearly endless situations to make sure its cars are prepared for the unexpected, which also happens to fit into the latest trend in the AI ​​space.

To be clear, Waymo’s world model is as meaningful as any use case for the technology. The company has a ton of high-definition data it’s collected from its time on the road that it can use to create realistic reconstructions of the roads. But, the company said, instead of building a model based only on that information, it’s going to use Google’s Genie 3 models to put its cars in simulated situations that extend far beyond what’s already in the data set, collected from cameras and lidar sensors.

We’re excited to introduce the Waymo World Model – a leading-edge generative mode for large-scale, hyper-realistic autonomous driving simulations. @GoogleDeepMindGenie of 3.

By simulating the “impossible”, we actively prepare Waymo drivers for some of the rarest and… pic.twitter.com/Pl80OMDqLC

– Waymo (@waymo) 6 February 2026

Google made a splash last month when it released a beta version of Genie 3 to the public, allowing a subgroup of paid customers to create 3D worlds with realistic physics. Unlike a large language model (LLM) – the underlying technology that powers most AI tools, including Google’s own Gemini – which use vast amounts of training data given to them to predict the most likely next part of the sequence, world models are trained on real-world dynamics, including physics and spatial properties, to create a simulation of how a physical environment operates.

genie 3 🤝 @waymo

The Waymo World Model generates photorealistic, interactive environments to train autonomous vehicles.

This helps cars deal with rare, unexpected events before they encounter them in reality. 🧵 pic.twitter.com/m6rlmkMFJH

– Google DeepMind (@GoogleDeepMind) 6 February 2026

Waymo plans to put its cars through scenarios they won’t find themselves in until it’s too late. This includes extreme weather conditions and natural disasters, so cars can figure out how to deal with tornadoes or flood waters; Sudden safety emergencies such as falling tree branches or an accident with a lot of debris; And like an elephant on the road, running into the unexpected. “By simulating the ‘impossible,’ we proactively prepare Waymo drivers for some of the most rare and complex scenarios,” the company said.

The theory is certainly correct, although world models are not without their shortcomings. Initial reaction to the consumer version of Genie 3 was a bit ambivalent, and world models are still susceptible to hallucinations. We’re still in the early stages of seeing these models deployed, and there’s a lot of room to iterate on them.

And Wemos certainly has its issues in real-world edge-case scenarios. Late last year, a Waymo crushed an adorable Bogda cat named Kit Kat, and last month, it collided with a child in a school zone. Those interactions aren’t particularly rare for the driver, so hopefully Waymo can refine its responses in those scenarios in addition to preparing for the most unexpected situations.