Home / Technology / AI Tests AI: Laplace Builds Safer Robotics
AI Tests AI: Laplace Builds Safer Robotics
13 Oct
Summary
- Laplace uses AI to automatically discover failure points in physical AI systems.
- The company builds high-fidelity simulations grounded in real-world data.
- Laplace aims to ensure the safety of robots and autonomous vehicles before deployment.
Laplace Intelligence, an Ann Arbor-based startup, is developing AI-powered red-teaming solutions to ensure the safety of physical AI systems. Founded by Howie Sun and Henry Liu, the company utilizes AI agents to stress-test robots and autonomous vehicles, simulating real-world scenarios to uncover potential failure points.
The core of Laplace's technology involves a high-fidelity simulation environment reconstructed from real-world data. This platform, built upon Nvidia's Cosmos, allows for the creation of environments true to scale, with developers able to dynamically alter settings through language prompts.
Laplace's unique selling proposition lies in its AI agents, which autonomously design corner cases by analyzing diverse information sources, including news and collision reports. This approach moves beyond traditional methods of training AI with recorded footage or human-designed edge cases.
The company has already secured revenue and modest profits from autonomous driving clients like Isuzu and Bot Auto, with Nvidia also being a customer. Laplace is actively seeking to scale its operations and transition to recurring revenue models, positioning its AI-testing methodology as a new paradigm for physical AI safety.
