s world models gain momentum, Waymo might be putting this lofty tech to good use.
On Friday, the robotaxi firm unveiled the Waymo World Model, a generative model for “large-scale, hyper-realistic” autonomous driving simulation. Waymo’s model is built on Google's Genie 3, the latest iteration of its world model series, released in August.
The model’s architecture allows the engineers to better modify and control the scenes it generates using language, allowing for everything from time-of-day or weather condition changes to entire synthetic scene generation with simple language prompts. The model can also convert authentic dashcam videos into synthetic scenes for training.
With the Waymo World Model, the company leveraged Genie’s ability to generate photorealistic, interactive 3D environments, applying it specifically to the “driving domain.” This allows the world model to generate rare events, including:
- An elephant appearing in the middle of the road;
- Being chased down a freeway by a tornado;
- And driving through a tropical city that happens to be covered in snow.
This kind of training could help prepare Waymo to scale across new locations and driving environments, potentially even allowing the robotaxis to better handle natural disasters.
The company has faced scrutiny for its safety practices in recent weeks after one of its vehicles struck and injured a child in California. In a hearing this week, Waymo’s Chief Safety Officer Mauricio Peña acknowledged that the company relies on remote operators overseas for “guidance in certain situations,” but these operators do not pilot the vehicles themselves.
“By simulating the 'impossible', we proactively prepare the Waymo Driver for some of the most rare and complex scenarios,” The company said in its announcement.
Waymo’s model adds to growing enthusiasm for world models. Google last week released Project Genie, an “experimental research prototype” built on Genie 3 that lets users create and explore virtual worlds and video game-like environments. World model startups from AI pioneers like Fei-Fei Li and Yann LeCun are also in talks for massive funding rounds as investors search for the next big thing after LLMs.