Laplace Intelligence red-teams robots and self-driving cars before they hit the real world
Laplace Intelligence is developing automated red-team testing to find weaknesses in robots, self-driving cars and other physical AI systems before they are deployed. The company argues that testing in realistic simulations can help prevent unpredictable AI behaviour from causing problems in the real world.
Founded a year ago by Howie Sun and Henry Liu at the University of Michigan, Laplace builds on their research into testing autonomous driving systems in simulation, including a 2023 paper published in Nature. Its platform adapts Nvidia’s Cosmos to recreate real places using sources such as map data and Street View, then lets developers change conditions with prompts and test how systems respond. The article’s available text ends during a description of a demonstration.
- Laplace uses simulations to probe physical AI for failures.
- Its platform recreates real-world settings for safety tests.
- The founders’ work grew from university research.
New here? Start with this
Robots and self-driving cars use artificial intelligence to interpret their surroundings and decide what to do. Before these systems are used in public or workplaces, developers test them to find situations in which they might make unsafe or unexpected choices.
Laplace Intelligence was founded by Howie Sun and Henry Liu, researchers at the University of Michigan. Its work draws on research into testing self-driving systems in computer simulations, including a paper the founders published in 2023.
The company’s platform is designed to recreate real places in simulation and let developers alter conditions to see how a system responds. Such testing matters because physical AI systems can affect people and property, though simulated tests cannot cover every circumstance they may encounter in the real world.
AI Art Business Celebrity Companies Culture Entertainment Technology