Nvidia Unveils New AI Models and Tools for Robotics: Cosmos Reason, Transfer-2, and Beyond – Insights and Data

Nvidia has launched innovative AI models and infrastructure aimed at robotics developers, highlighted by the 7-billion-parameter “Cosmos Reason,” a vision-language model for physical AI. This suite enhances synthetic data generation and AI reasoning in robotics. Notable tools include “Cosmos Transfer-2,” which accelerates synthetic data generation from 3D simulations, and a distilled version for improved performance. Presenting at the SIGGRAPH conference, Nvidia emphasized these solutions for creating datasets vital for training robots and AI agents. Analysts predict an 18.2% CAGR for the global robotics market through 2030, driven by AI advancements. “Cosmos Reason” equips robots with reasoning capabilities by utilizing memory and physics principles, aiding in tasks such as video analytics. Additionally, Nvidia unveiled neural reconstruction libraries for 3D simulations and updates to its Omniverse SDK. New servers, including the RTX Pro Blackwell Server and DGX Cloud, are introduced to enhance robotics workflows, underscoring Nvidia’s strategic move into robotics as it seeks diversified applications for its AI GPUs.

Source link