Laplace intelligence uses AI to safeguard physical robots and self-driving cars before deployment

Laplace Intelligence, a startup based in Ann Arbor, Michigan, is leveraging AI to stress-test robotics and autonomous vehicles in simulation before they are deployed in the real world. Founded a year ago by Howie Sun and Henry Liu, both former University of Michigan researchers, the company aims to ensure the safety of physical AI systems by identifying and fixing potential failures. Sun, who previously spent eight years in academia researching autonomous driving AI, emphasized the importance of proactive testing, citing the OpenAI-Hugging Face incident as a cautionary example. Laplace’s approach involves high-fidelity simulations grounded in real-world data, powered by Nvidia’s Cosmos platform. These simulations, which can recreate environments like Arizona highways or Paris pedestrian crossings, are enhanced with real-time data from sources like TomTom. The startup’s AI agents are designed to automatically generate edge cases, challenging autonomous systems in ways that human designers might miss. Laplace is currently working with companies like Bot Auto and a major Japanese automaker, as well as Nvidia and T-Mobile on a smart city pilot in San Jose, California. Competing against firms like Applied Intuition, Laplace’s CEO Sun believes the company’s AI-driven testing methodology represents a new paradigm in physical AI safety, offering a dynamic and interactive approach to validation.