Laplace Intelligence, a startup born from academic research at the University of Michigan, employs automated AI-driven red-teaming to uncover weaknesses in robots and self-driving vehicles within high-fidelity simulations, aiming to enhance physical AI safety before public use.
- Automated AI agents simulate rare, risky scenarios to stress-test physical AI.
- Simulations use real-world data and interactive actor behaviors for accuracy.
- Laplace competes at TechCrunch Disrupt showcasing its safety validation platform.
What happened
Laplace Intelligence, a University of Michigan spinout founded by Howie Sun and Henry Liu, has developed an advanced simulation platform to stress-test robots and autonomous vehicles before deployment. The startup leverages AI agents that actively explore and challenge physical AI systems within high-fidelity simulations that accurately reconstruct real-world environments.
The simulation environment is based on a refined version of Nvidia’s Cosmos platform, enhanced with real-world map data and live traffic inputs from providers like TomTom. This allows Laplace to recreate realistic settings ranging from highways to city streets, where AI-driven participants dynamically respond to conditions, providing an interactive and comprehensive testing ground.
Why it matters
As AI systems increasingly operate in the physical world, unpredictable failures could have serious consequences. Laplace’s approach addresses this safety challenge by applying automated red-teaming methods—traditionally used in cybersecurity—to physical AI. This process identifies corner cases and edge scenarios that pose the highest risk before a product encounters the complexities of the real world.
The startup’s technology offers a critical layer of assurance for companies, insurers, and regulators by providing evidence-backed validation of autonomous systems’ reliability. By reducing reliance on manual testing and enabling continuous discovery of failure points, Laplace aims to enhance trust and safety in emerging robotic and autonomous vehicle technologies.
What to watch next
Laplace Intelligence will be presenting its technology at TechCrunch Disrupt’s Startup Battlefield 200, a prestigious startup competition, signaling growing investor and industry interest in AI safety for autonomous systems. Observers should watch for potential partnerships or pilot programs with autonomous vehicle manufacturers and robotics firms.
Additionally, the company’s ongoing development of AI agents that generate challenging and rare scenarios based on real-world data like news and police reports could expand its applicability across other sectors deploying physical AI, such as smart cities and industrial automation. Market adoption and regulatory engagement will be key indicators of Laplace’s impact going forward.