Laplace Intelligence, based in Ann Arbor, is developing automated red-teaming tools for robotics and self-driving cars to ensure safety before real-world deployment. The company aims to help companies discover and fix problems before deployment, building a shared foundation of evidence that customers, insurers, and regulators can trust.
The startup reported that its high-fidelity simulation environment is grounded in the real world, allowing it to reconstruct environments like Arizona highways and Paris pedestrian crossings. This simulation is built using real-world map data and can be modified in real time with language prompts to adjust settings like traffic conditions.
Laplace's technology is built on Nvidia's Cosmos platform, which is post-trained to create realistic environments. The company's AI agents are designed to automatically design corner cases, challenging autonomous vehicles and other physical AI agents by drawing on news, police reports, and other data sources.
"We are essentially using simulation to predict the real world’s future," said Howie Sun, CEO of Laplace Intelligence. Sun explained that the company's goal is to ensure that any physical AI is safe in simulation before it is deployed in the real world.
The announcement follows Laplace's participation in Startup Battlefield 2, a pitch competition at TechCrunch Disrupt in San Francisco. The company is also working with partners building humanoid robotics for warehouses and is in the early stages of developing recurring revenue models.
Laplace Intelligence did not say how it will handle the challenge of generalized robotics, where edge cases are less predictable. The company is working toward raising funds and scaling to compete with its rivals, including Applied Intuition.
Source: techcrunch
