KAIST will invest 13.7 billion won ($9.8 million) to build an artificial intelligence-based robotics innovation ecosystem and develop viable business models. The institute is also launching its deep tech scale-up valley initiative in earnest.
KAIST announced on the 27th that it held a "Deep Tech Scale-Up Valley Project Progress Report" at its main campus in Daejeon the previous day, unveiling its physical AI strategy and execution structure. Physical AI refers to technology that combines robotics and AI to enable autonomous judgment and action in the real world.
The Deep Tech Scale-Up Valley development project is a joint initiative by the Ministry of Science and ICT, Daejeon Metropolitan City, and KAIST, aiming to commercialize KAIST's deep tech robotics technologies and build a robotics innovation ecosystem. The project will proceed in phases, including establishing a physical AI platform, discovering and investing in startups, building demonstration testbeds, and expanding cooperation networks with global robotics companies.
KAIST has formed a robotics alliance with participants including KAIST Holdings, Daejeon Techno Park, Daejeon Center for Creative Economy and Innovation, Angel Robotics, and Eurobotics. Starting last year, the institute plans to invest a total of 13.65 billion won over three and a half years.
The project aims to create a virtuous cycle ecosystem and nurture future unicorn companies based on a three-pillar system of technology commercialization, deep tech R&D, and commercialization scale-up. In its first year last year, the project achieved results of 23 billion won in technology transfers and investment attraction through physical AI lectures, startup pitching, and investment networking.
Notably, this report redefined physical AI not as a simple AI technology competition but as "an industrial structure problem." The argument is that commercialization becomes difficult unless R&D, industrial sites, and investment ecosystems are organically connected.
KAIST plans to establish a clear value chain connecting researchers, industry experts, and companies. Through this, the strategy is to expand physical AI beyond laboratory demonstration levels to technology that can be applied to actual industrial sites and solve problems.
At the report, opinions were also presented that "for physical AI to be applied to industrial sites, 'meaningful data' generated from actual industrial sites is needed, beyond virtual environments." The explanation is that data reflecting physical senses and judgment must be accumulated through collaboration with experts in manufacturing processes, and an execution system must be built where robots can continuously cooperate without interfering with experts' work.
The need to actively develop technology that enables robots trained in virtual environments to perform equally well in reality was also emphasized. Professor Kong Kyoung-chul of KAIST's Department of Mechanical Engineering said, "It is now important to organize the mixed concepts of physical AI and create specific platforms that anyone can use." He added, "For AI learned in virtual environments to work well with actual robots in real environments, not only must the technological accuracy of the virtual world improve, but physical variables in the real world must also be predictably and stably managed."