AI World Models Set to Transform Robotics and Autonomous Driving
AI-generated image · US National Wire
Companies are leveraging 3D simulations to train AI systems on real-world physics, as reported by Wired.
AI companies are increasingly developing "world models" to simulate real-world physics, a technology that as Wired first reported is poised to accelerate the expansion of home robotics and driverless cars by 2027. Unlike text-based large language models, these systems use visual inputs to predict future visual states, creating digital training grounds for AI.
Industry leaders are heavily investing in the sector. Niantic Spatial is utilizing environmental data to build replicas of the real world, while Odyssey CEO Oliver Cameron notes that these simulations allow robots to practice in virtual environments before entering physical spaces. Similarly, Wayve is using dashcam footage to recreate hazardous road conditions for autonomous vehicle training.
The field has seen significant financial movement, including over $1 billion raised by Yann LeCun, Meta's former chief AI scientist, for his startup Advanced Machines Intelligence. Additionally, AMD acquired World Labs—a startup launched in 2024 by Stanford computer scientist Fei-Fei Li—for more than $8 billion.