SKAI WorldWide and SKAI Intelligence Showcase Physical AI Synthetic Data Technology on JTBC Documentary
[Edaily]
AI agent and database (DB) solutions company SKAI WorldWide (KOSDAQ: 357880) announced on the 24th that the physical AI industrial synthetic data generation technology developed by its affiliate, SKAI Intelligence, was featured in JTBC’s special documentary “Korea’s Winning Strategy – Physical AI.”
Aired on the 23rd, the documentary examined Korea’s competitiveness in the era of physical AI, focusing on the country’s capabilities in semiconductor manufacturing, robotics, and software development. In particular, it highlighted the challenge of securing the training data required for the commercialization of physical AI.
The perception and performance of physical AI systems and robots can vary depending on changes in surrounding conditions, such as lighting and object positions. Since training robots to respond to all possible variables in real-world environments requires significant time and cost, synthetic data technology, which generates training data in virtual environments, is emerging as an alternative.
The program introduced SKAI Intelligence’s synthetic data generation technology and infrastructure, as well as a cable routing project being conducted jointly with Seoul National University. Cables are considered a challenging application because they are flexible objects whose shapes continuously change, requiring robots to perceive their surroundings and perform tasks accordingly.
SKAI Intelligence has developed technology based on digital twin technology that represents physical factors such as friction coefficients and lighting conditions as mathematical parameters in virtual environments, enabling the generation of training data that allows physical AI systems to learn in 3D and 6D spaces.
The company can also reproduce scenarios that are difficult to capture with sufficient real-world data, such as components with scratches, contamination, or deformation, within virtual environments. It is also building technology based on NVIDIA Omniverse to reproduce materials, lighting conditions, and camera settings found in industrial environments while automating the entire process from synthetic data generation to data labeling.
Through these technologies, the company plans to establish a “Real-to-Sim-to-Real” data loop, connecting real-world data to virtual environments and ultimately bringing the results back into real-world applications.
In the documentary, Lee Jae-chul, CEO of SKAI Intelligence, said, “We call it training synthetic data when data can be calculated, tracked, and reused, and can be delivered directly to physical AI systems for immediate training.” He added, “What matters is not simply generating data, but building data infrastructure that creates an entire data loop to address the bottlenecks facing physical AI.”
Shin Jae-hyuk, CEO of SKAI WorldWide, said, “The true asset of Korean manufacturing lies in the know-how embedded in the hands of skilled workers. When experienced workers retire, that knowledge can be lost,” adding, “It is important to turn that know-how into data and pass it on to robots so that it can be inherited and sustained across the industry.”
SKAI WorldWide operates an AI agent and database solutions business based on graph database (GDB) technology. Its affiliate, SKAI Intelligence, has signed a framework agreement with ABB Robotics and is conducting joint research on robot vision technology with the AI Institute of Seoul National University (AIIS).