Laplace red-teams robots and self-driving cars with AI — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Laplace Intelligence builds automated red-teaming tools that use AI agents to probe robots and self-driving cars for failure points in high-fidelity simulation before real-world deployment.
- Co-founders Howie Sun (CEO) and Henry Liu launched the Ann Arbor-based startup roughly a year ago after their 2023 Nature paper on challenging autonomous vehicle AI drew interest from AV companies; Sun had spent eight years on the research.
- The platform runs on a post-trained version of Nvidia's Cosmos world simulation, reconstructing real-world environments — from Arizona highways to Paris pedestrian crossings — using Google Street View imagery and real-time traffic data from TomTom.
- Laplace's differentiator is AI agents that autonomously design corner cases by mining news articles, police collision reports, and other sources to challenge physical AI, aiming to move humans further out of the test-creation loop.
- Revenue is flowing from autonomous trucking firm Bot Auto, a major Japanese automaker, and humanoid robotics warehouse projects, alongside a free San Jose smart-city traffic pilot with Nvidia and T-Mobile; the company is already modestly profitable.
- Chief rival Applied Intuition has raised $1.2 billion at a $15 billion post-money valuation; Laplace is competing in TechCrunch Disrupt's Startup Battlefield 200 in San Francisco (Oct 13-15) to raise funds and scale.
Why it matters: With AI agents moving from computers into the physical world, proving safety before deployment is becoming a gating problem — and capital is already following: Applied Intuition has raised $1.2 billion at a $15 billion valuation on a related testing thesis, meaning even a Disrupt-stage startup like Laplace is entering a cash-heavy race where the validation of AI-stress-testing-AI has already drawn billion-dollar backing.
Ask SkimNews
