FamCheck. Everyone who looks after your kids stays in the loop. $9.99 a month for the whole household, first week free, no card. famcheck.coSPONSORED
The Miami Art Journal

Tech

Laplace Intelligence stress-tests physical AI systems

The Editors ·

Laplace Intelligence stress-tests physical AI systems

A report this week details how Laplace Intelligence is developing automated red-teaming protocols for robotics and self-driving cars. The company aims to identify vulnerabilities in physical AI before these systems enter the real world, drawing a parallel to the unpredictable nature of digital AI agents. The approach involves deliberately attempting to make these systems fail within simulation environments to uncover potential issues.

Howie Sun, the CEO and co-founder of Laplace Intelligence, notes that as AI agents move from computers into the physical world, ensuring their safety becomes paramount. The company is betting on the premise that AI should test physical AI to guarantee reliability. Sun suggests that this method helps companies discover and fix problems prior to deployment, building a foundation of evidence for customers, insurers, and regulators.

Sun and his co-founder and chief scientist Henry Liu established the company a year ago after their research on challenging autonomous driving AI in simulation gained attention. Their 2023 paper on the subject appeared on the cover of Nature, leading to inquiries from autonomous vehicle companies interested in applying their academic findings to practical products. The name Laplace references a thought experiment by Pierre-Simon Laplace regarding the prediction of the future through precise knowledge of the universe’s state.

The startup is currently competing in the Startup Battlefield 200 at the report Disrupt in San Francisco. Laplace is scheduled to present on the Disrupt Stage on Tuesday afternoon, October 13. The event features twenty startups, and Laplace is one of the participants showcasing its technology to the startup community.

The company’s methodology relies on simulation to predict outcomes in the real world, a concept rooted in the academic work of Sun and Liu. Sun spent eight years in academia researching how to challenge autonomous driving AI, which eventually led to the creation of a product designed for the real world. The goal remains to provide a shared foundation of trust for those interacting with physical AI systems.

Source: TechCrunch AI