What Is Waymo World Model — How Does It Help Train Driverless Cars?

The system, dubbed the Waymo World Model, aims to train Waymo Driver on rare and complex driving situations before the vehicles encounter them in real traffic.

• The world model builds photorealistic 3D environments using advanced AI to test autonomous behavior under extreme scenarios.
• By leveraging Genie’s rich understanding of the physical world, Waymo’s model produces high-fidelity sensor outputs, including detailed camera and lidar imagery. 
• Waymo’s flagship autonomous driving system has already accumulated nearly 200 million fully driverless miles across U.S.cities.

Waymo, the self-driving technology unit of Alphabet Inc. (GOOG, GOOGL), has introduced a simulation platform designed to advance autonomous vehicle testing and safety. 

The system, dubbed the Waymo World Model, aims to train Waymo Driver on rare and complex driving situations before the vehicles encounter them in real traffic.

Waymo World Model 

The company announced the new generative simulation model, detailing how it builds photorealistic 3D environments using advanced AI to test autonomous behavior under extreme scenarios and edge-case events, and positioning the tech as a critical component of its safety-first AI strategy.

The World Model builds on Google DeepMind’s Genie 3, a powerful world simulator capable of generating realistic virtual scenarios such as tornadoes, floods, and unexpected animal crossings.

Alphabet’s class A stock traded over 2% lower on Friday afternoon. On Stocktwits, retail sentiment around the stock remained in ‘extremely bullish’ territory amid ‘extremely high’ message volume levels. 

Hyper-Realistic, Multi-Sensor Simulation

Waymo’s flagship autonomous driving system has already accumulated nearly 200 million fully driverless miles across U.S. cities, with simulation used to supplement real-world data. 

By leveraging Genie’s rich understanding of the physical world, Waymo’s model produces high-fidelity sensor outputs, including detailed camera and lidar imagery. Engineers can control virtually every aspect of the simulation, from road design and traffic actions to weather, using simple language prompts.

Traditional simulators rely heavily on past driving data collected from deployed fleets, but the Waymo World Model enables testing of scenarios that rarely occur in real life.

On Monday, Waymo said it secured $16 billion in fresh funding, valuing it at $126 billion. The company also said its self-driving cars have completed 127 million fully autonomous miles and cut serious injury crashes by 90% compared with human drivers.

For updates and corrections, email newsroom[at]stocktwits[dot]com.<