Waymo is leveraging Genie 3, a world model developed by Google DeepMind, to enhance its self-driving car technology. The new model creates "hyper-realistic" simulated environments, allowing Waymo to train its AI on rare and challenging scenarios. This includes situations like driving in snow or navigating complex urban environments, which are often underrepresented in real-world training data.

This marks a significant step forward from relying solely on real-world driving data. Waymo's use of Genie 3 allows for the creation of simulations using simple prompts and driving inputs. The model generates both 2D video and 3D lidar outputs, providing the AI with rich data for improved perception and decision-making on the road.

By simulating diverse conditions, Waymo aims to improve its self-driving capabilities, especially in new markets with challenging weather. The technology enables the company to test how its vehicles would respond to various situations. With the ability to mutate conditions in existing videos, Waymo can refine its AI's ability to handle unexpected events.

For consumers, this means safer and more reliable autonomous vehicles. The ability to simulate and train for a wider range of scenarios is crucial for the future of self-driving technology. Waymo’s expansion into new cities like Boston and Washington, D.C., will benefit from the enhanced capabilities of the Waymo World Model.