엔비디아가 찾아왔다 한국식으로 로봇 가르친다 (스카이인텔리전스 이재철 대표)

※ This video was produced with the support of Sky Intelligence. 💡 Content Summary Physical AI, which enables robots to perform physical tasks like humans, faces a fundamental problem regarding data acquisition. Unlike language or image-based AI, robots must learn physical information such as friction, reflection, and assembly deviations; however, collecting this data directly in the real world is excessively costly and time-consuming. In this video, Lee Jae-cheol, CEO of Sky Intelligence, appears to explain the method used to resolve this bottleneck. Sky Intelligence adopts a method where robots are first trained within a digital twin that replicates an actual factory. After applying the results to the real world, if errors are discovered, the data is returned to the virtual environment to make corrections. This technological capability, which generates synthetic data based on the NVIDIA Omniverse platform, has been externally recognized, as Sky Intelligence was named a key contributor in the Physical AI white paper jointly published by ABB Robotics and NVIDIA. SoftBank Chairman Masayoshi Son predicted that Physical AI will account for 20% of global GDP by 2030. The video also projected that, given Korea, China, and Japan together make up half of the world's manufacturing sector, Korean companies have just over three years to seize the lead in this field. How will the growth of the Physical AI industry and opportunities for Korea unfold? 📌Key Topics • Physical AI Data Bottlenecks • Digital Twin Training Methods • NVIDIA Omniverse Collaboration • Participation in the Physical AI White Paper • Manufacturing Opportunities in Korea, China, and Japan [This is an AI-generated summary of the video] 👉Understanding Inquiries: [email protected] 👉 Read "Understanding Text". https://contents.premium.naver.com/ba... 👉 Join Understanding Membership    / @understanding.