Laplace Intelligence Automated Red-Teaming Stress-Tests Physical AI Before Real-World Deployment
· Technology · TechCrunch
Laplace Intelligence, an Ann Arbor startup founded by former University of Michigan researchers Howie Sun and Henry Liu, builds automated red-teaming to stress-test robotics and autonomous vehicle AI in high-fidelity simulation before real-world deployment. The company uses a post-trained version of Nvidia's Cosmos world simulation platform grounded in real-world data to reconstruct environments from Arizona highways to Indian roads, with AI agents deliberately probing for failure points. Laplace is one of 20 startups competing in TechCrunch Disrupt's Startup Battlefield, where it will present on the stage in downtown San Francisco on October 13. Sun spent eight years researching how to challenge autonomous driving AI in simulation, publishing a cover story in Nature in 2023 that attracted interest from autonomous vehicle companies. The red-teaming approach aims to discover and fix problems before deployment, eventually building a shared foundation of evidence that customers, insurers,
Why it matters
The red-teaming service directly addresses safety concerns for companies deploying physical AI like self-driving cars and robots, offering a simulated testing ground that could reduce real-world failures and influence future regulatory standards.
Read the original report — TechCrunch
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.