Nvidia Wants to Use AI in the Real World, From Biotech to Transportation

Core Insights - Nvidia is intensifying its efforts in the AI sector by releasing over 70 research papers that demonstrate the technology's capabilities in real-world applications beyond just text and images [1] Group 1: AI Advancements - The company is promoting "embodied intelligence," which refers to AI that can perceive, reason, and act across various industries such as manufacturing, biotechnology, and transportation [2] - Nvidia emphasizes that for AI to be effective, it must engage with real-world use cases, as stated by Bryan Catanzaro, vice president of applied deep learning [3] Group 2: Key Developments - A notable development is the Skill Reuse via Skill Adaptation (SRSA) system, which allows robotic agents to adapt previously learned skills for unfamiliar tasks, improving task success by 19% and reducing training sample needs by over 50% [4] - In biotechnology, Nvidia's Proteína model, which trains on 21 million synthetic protein structures, outperforms Google's DeepMind's Genie 2 in accuracy and diversity, potentially accelerating vaccine development and enzyme design [5] Group 3: Technological Innovations - The STORM (Spatio-Temporal Occupancy Reconstruction Machine) creates 3D maps in under 200 milliseconds, suitable for drones, AR systems, and autonomous vehicles [6] - The Nemotron-MIND system teaches large language models to solve math problems using synthetic dialogue, achieving better performance on key benchmarks while utilizing fewer tokens [6] Group 4: Conference Announcements - These advancements were presented at the 2025 International Conference on Learning Representations (ICLR), where Nvidia also introduced Inference Microservices (NIM), a platform designed to facilitate the deployment of advanced AI models without requiring extensive infrastructure [7]