Government launches industry-academia-research collaboration to industrialize physical AI
The government is set to fully launch efforts to industrialize 'physical AI,' which combines artificial intelligence (AI) with physical machinery such as robots, enabling them to perceive, judge, and act in the real world. Ryu Je-myung, Second Vice Minister of Science and ICT, emphasized the importance of industry-academia-research collaboration at 'Physical AI Tech Day,' held on the 14th at COEX in Gangnam, Seoul.
In his opening remarks, Vice Minister Ryu said, "Physical AI is a pillar of the three megaprojects—alongside semiconductors and AI data centers—that will lead Korea's future." He added, "Since it cannot be achieved by any single technology or company alone, close collaboration among industry, academia, and research institutions is essential." He continued, "We will actively support industrial application and expansion so that excellent technologies do not remain in research and development but lead to field validation and commercialization."
Vice Minister Ryu also toured the technology exhibition at the venue, examining technologies showcased by domestic companies and research institutions—including AI semiconductors and data, AI-based factory optimization, collaborative robots, and humanoids—and assessed the potential for applying R&D achievements in industrial settings.
The keynote speeches featured Jung Byung-kon, CEO of Friendly AI; Kim Sang-bae, professor at the Massachusetts Institute of Technology (MIT); and Lee Kyu-bin, professor at the Gwangju Institute of Science and Technology (GIST). CEO Jung identified frontier AI model development for robot brains and inference efficiency (inference infrastructure) as key challenges, proposing methods to reduce recomputation through KV cache reuse and cache-aware routing. Professor Kim said intelligence architectures must be explored to increase robot task success rates in physical environments. Professor Lee suggested that robot foundation models should be able to output both language and motion.
