To make autonomous driving safer and more reliable, Waymo (GOOG) (GOOGL) is using DeepMind’s Genie 3 AI model to create realistic digital worlds to help its autonomous vehicles handle hard-to-predict situations and expedite the global rollout of its robotaxis.
The Waymo World Model, built on Genie 3, generates photorealistic and interactive 3D environments and “adapts for the rigors of the driving domain.” By using Genie’s “immense” world knowledge, Waymo (GOOG) (GOOGL) can be trained to react to situations such as extreme weather conditions and natural disasters.
“Genie 3’s strong world knowledge, gained from its pre-training on an extremely large and diverse set to videos, allows us to explore situations that were never directly observed by our fleet,” Waymo said in a blog post.
Waymo (GOOG) (GOOGL) is then incorporating that data into its 3D lidar outputs that are unique to its hardware suite. This allows for more comprehensive visual details that incorporate depth, the behavior of other uses of the road, as well as custom mutations.
“The Waymo World Model can generate virtually any scene – from regular, day-to-day driving to rare, long-tail scenarios—across multiple senor modalities,” the company said.