Innoviz Paper Explores 'Physical AI'

- Physical AI refers to artificial intelligence systems that can perceive, reason, and interact with the physical world, unlike generative AI models like GPT that operate on digital text and images. These systems combine AI algorithms with sensors and actuators to allow machines like robots and autonomous vehicles to perform complex actions in real-world environments. - The paper's co-author, Innoviz CEO Omer Keilaf, previously served as an officer in an elite technological unit of the Israeli Defense Forces and held senior leadership roles at STMicroelectronics and Consumer Physics before co-founding Innoviz. - The "World Models" mentioned in the paper are AI systems trained on vast amounts of real-world data to understand the dynamics of the physical world, including geometry, motion, and physics. These models can create realistic simulations, often called digital twins, to train and test Physical AI systems before they are deployed. - The white paper argues that the primary bottleneck for deploying Physical AI is no longer processing power, but access to high-quality, real-time 3D data. It posits that while cameras infer depth and radar lacks resolution, LiDAR directly measures distance to create precise geometric data. - Beyond autonomous vehicles, Innoviz is deploying its LiDAR technology for "smart infrastructure." For instance, its InnovizSMART system has been deployed at critical infrastructure sites in Israel for perimeter security, using its 3D sensing to detect and classify objects with high precision. - The company has a major partnership with Daimler Truck and Torc Robotics to supply its LiDAR sensors for the series production of SAE Level 4 autonomous trucks. - Innoviz announced 2025 revenue guidance of $50-60 million, which is more than double its 202