What Happened

Laplace Intelligence, a University of Michigan spinout, is using AI agents and interactive simulation to stress-test robots and self-driving vehicles before they hit the real world. The startup is competing in Startup Battlefield 200 at TechCrunch Disrupt in San Francisco, October 13-15.

The Key Details

  • Laplace builds automated red-teaming for robotics and physical AI - essentially stress-testing systems by deliberately trying to make them fail
  • The company's core product is a high-fidelity simulation reconstructed from the real world, paired with AI agents that probe for weaknesses and failure points
  • The simulation environment is based on Nvidia's Cosmos, an AI world simulation platform, grounded in real-world data
  • Developers can use language prompts to change simulation settings in real time, such as adding afternoon traffic from TomTom data
  • Within simulations, both the environment and behavior of other actors are interactive - other entities react to the test agent
  • The unique selling point is AI agents that automatically design "corner cases" - tricky, rare scenarios that could trip up AI systems
  • Laplace is generating revenue, though modest profit, with customers including Japanese trucking manufacturer Isuzu, Texas-based autonomous trucking firm Bot Auto, and Nvidia
  • The startup is also working with partners building humanoid robotics for warehouses and T-Mobile on a smart city pilot for San Jose, California
  • Chief rival is Applied Intuition, which has raised $1.2 billion at a $15 billion post-money valuation
  • Laplace was founded a year ago by Howie Sun and Henry Liu, formerly of University of Michigan
  • In 2023, Sun and Liu published a paper on challenging autonomous driving AI in simulation that appeared on the cover of Nature

Why It Matters

  • As AI agents move from computers into physical systems, ensuring their safety before deployment becomes critical
  • Laplace's approach addresses the limitation of current methods that rely on recorded footage or human-designed edge cases
  • The technology could be particularly important as generalized robots are placed in real-world environments where edge cases are less documented
  • By using AI to test AI in simulation, Laplace aims to create a shared foundation of evidence that customers, insurers, and regulators can trust
  • The need for this kind of testing will become more apparent as generalized robots enter real-world use

What It Means for Readers

  • For technology enthusiasts: demonstrates an emerging approach to making physical AI safer through simulation-based testing
  • For autonomous vehicle developers: offers an alternative to relying solely on recorded footage or human-designed edge cases
  • For businesses and investors: shows a growing market for AI safety testing in physical systems, with Laplace already generating revenue
  • For everyday users: highlights the behind-the-scenes work to make self-driving cars and robots safer before they appear on roads and in public spaces

What Happens Next

  • Laplace plans to raise funds and scale following its Disrupt competition participation
  • The startup is working toward turning project-based contracts into recurring revenue
  • As generalized robots enter real-world use, demand for this type of testing technology is likely to grow
  • Laplace will compete with rivals like Applied Intuition in the physical AI testing space

VANTIX Takeaway

Laplace Intelligence is applying AI-driven simulation to make physical AI safer before deployment - a approach that could become increasingly important as robots and self-driving vehicles move from controlled testing into real-world use.