Laplace Intelligence uses AI agents to red-team robots and self-driving cars

Simulation grounded in the real world

Howie Sun, CEO and co-founder of Laplace Intelligence, said AI agents are moving from computers into the physical world, and the company wants failures to be found before deployment. Laplace Intelligence is based in Ann Arbor and was founded about a year before TechCrunch Disrupt by Sun and chief scientist Henry Liu out of the University of Michigan.

Sun spent eight years in academia researching how to challenge autonomous driving AI in simulation. In 2023, Sun and Liu, his then PhD adviser, published a paper on the subject that appeared on the cover of Nature. Sun said calls from autonomous vehicle companies followed, which led him to turn the research into a product.

A test world that reacts

Laplace says its core product pairs a high fidelity simulation rebuilt from the real world with an AI agent that probes for weaknesses and failure points. The simulation is a post trained version of Nvidia Cosmos, an AI world simulation platform, and is grounded in real world information so environments such as an Arizona highway or a Paris pedestrian crossing can be reconstructed.

The system blends a mesh of real world information. A Google Street View image can serve as a reference for generating a street, while real world map data keeps the result true to scale. Developers can use language prompts to change the setting in real time, and the company says it pulls real time traffic and congestion information from navigation providers such as TomTom.

AI agents design corner cases

Sun said the interactive part matters. In a demo, he directed the test agent vehicle to swerve into oncoming traffic, and the other vehicle moved out of the way instead of being driven through. The company frames the result as a world model with interactive participants rather than a fixed playback of scenery.

Laplace is also developing AI agents that automatically design corner cases, the rare scenarios that can trip up an AI system. Sun said the agents can draw on information such as news and police reports on collisions to challenge autonomous vehicles and other physical AI agents. He argued this moves humans further out of the loop compared with relying on recorded footage turned into simulation or on people inventing new scenarios.

Customers and next steps

The company says it has generated revenue and turned a modest profit, mostly from autonomous driving work. Customers named by Sun include Japanese trucking manufacturer Isuzu, Texas based autonomous trucking firm Bot Auto, and Nvidia. Laplace is working with Nvidia and T-Mobile on a free smart city pilot for San Jose, California, for high level traffic operations, and it is in early discussions with partners building humanoid robotics for warehouses.

Laplace is one of 20 startups in Startup Battlefield 200 at TechCrunch Disrupt in San Francisco, where it plans to present on the Disrupt Stage on Tuesday afternoon. Most contracts are project based, and the company says it is moving toward recurring revenue while seeking funds to scale. Sun named Applied Intuition as a chief rival and said Laplace moat is its methodology of using AI to test AI.

Sources: Techcrunch