Laplace Intelligence stress-tests robots with AI
University of Michigan spinout Laplace Intelligence is bringing automated AI-driven red-teaming to robotics and autonomous vehicles to uncover critical safety flaws before real-world deployment.
Ann Arbor-based startup Laplace Intelligence has developed an automated red-teaming platform designed to stress-test physical AI systems like autonomous vehicles and warehouse robotics in virtual environments. Founded a year ago out of the University of Michigan by CEO Howie Sun and Chief Scientist Henry Liu, the company commercializes eight years of academic research on challenging self-driving models. Their foundational methodology hit the cover of the journal Nature in 2023, drawing interest from autonomous vehicle operators looking to discover failure points safely. The startup's name references Pierre-Simon Laplace and his 1814 thought experiment, "Laplace's demon," which posited that complete knowledge of a current state allows perfect prediction of the future.
Laplace builds high-fidelity, interactive simulations using a post-trained version of Nvidia's Cosmos platform combined with Google Street View imagery and real-time traffic data from navigation vendors like TomTom. Instead of relying on static footage or manual edge-case design, Laplace deploys autonomous AI agents that actively probe physical AI systems for weaknesses. These agents automatically construct tricky corner cases drawn from police reports, news stories, and collision records. Within these simulations, background actors dynamically react to aggressive actions, such as an oncoming vehicle swerving to avoid a crash, creating an interactive testing environment.
The company is already profitable on project-based contracts while competing in Startup Battlefield 200 at TechCrunch Disrupt in San Francisco from October 13 to 15. Laplace counts Japanese vehicle manufacturer Isuzu and Texas-based autonomous trucking firm Bot Auto among its clients. Additionally, the startup is collaborating with Nvidia and T-Mobile on a free smart city pilot in San Jose, California, while expanding into humanoid robotics for warehouse management.
As Laplace transitions toward a recurring revenue model and seeks fresh investment, it aims to challenge established competitors like Applied Intuition, which has raised $1.2 billion at a $15 billion post-money valuation. While Applied Intuition also counts Isuzu as a client, Laplace aims to differentiate itself through its dynamic, agent-driven methodology that uses generative AI to continuously challenge real-world robotic deployments.
This is our own summary of reporting by TechCrunch AI
