Dennis Hong: South Korea Must Forge Its Own Path in Physical AI

Dennis Hong, a professor at UCLA and director of the RoMeLa research lab, outlined a distinct survival strategy for South Korea amid the escalating competition in 'Physical AI' led by the United States and China. He emphasized that South Korea cannot simply replicate the approaches of these two nations, which have strong ecosystems and capital in AI and hardware, respectively.

In an interview with Aju Economy on August 29, Hong stated, "It will become increasingly difficult to simply categorize the U.S. as AI (software) and China as hardware." He highlighted the importance of iteration speed in Physical AI, which involves creating and deploying robots to gather data, learn, and improve. He noted that the key factor is not who creates the best design but who can evolve the technology the fastest.

Hong identified South Korea's manufacturing base as a unique strength. He remarked, "South Korea has a solid foundation in semiconductors, automobiles, batteries, electronics, and precision manufacturing." He stressed that South Korea should not engage in a numbers game by trying to produce the same humanoids as the U.S. and China. Instead, the country should focus on integrating its globally competitive industries with Physical AI to excel in specific fields.

He particularly pointed out the significance of 'on-site data' already possessed by major South Korean companies like Samsung, Hyundai, and LG. While generative AI relies on data sourced from the internet, the critical data for Physical AI is found in real-world applications. He stated, "Not every company needs to create a general-purpose humanoid. We should leverage the strengths of large corporations that already have products, factories, supply chains, and actual usage scenarios to identify tasks for robots, gather data, learn, and establish a cycle of improvement."

However, he also mentioned the need to address 'social trust' as a challenge in combining manufacturing data with Physical AI. Concerns about job loss and data ownership disputes must be proactively resolved during the process of training AI with workers' skilled knowledge.

Hong noted, "Workers may think, 'If I teach my skills to AI, will I lose my job?'" He emphasized the necessity for a framework to clarify data ownership, consent, and how the value generated from that data will be shared.

Additionally, he cautioned against the exaggerated hype surrounding the technological advancements in Physical AI. He explained, "Achieving success in a pilot program is entirely different from deploying a system reliably in the field all day long." He added that predicting how many humanoids will be deployed in a few years is challenging in an environment where technological progress and hype coexist. He concluded, "The most significant change will be robots evolving from mere programmed machines to machines that learn and adapt. It is more important to demonstrate reliability and economic value in real-world applications."

Known as the 'Leonardo da Vinci' of robotics, Hong shared his research philosophy of aiming for creative, innovative, yet warm robots. He stated, "The most important value of robots is not that they work cheaper than humans, but that they take on tasks that are dangerous for people." He explained that robots should replace humans in hazardous environments such as disasters, fires, nuclear accidents, and toxic settings. He added, "The warmth should come not from the robot's heart but from the heart of the person who creates it, and it is crucial to consider for whom and for what purpose the technology is being developed."