"The person who claims to know precisely the future of artificial intelligence (AI) is indeed dangerous."
During the panel discussion of 'SMARTCLOUD SHOW 2025' held at the Westin Josun Hotel in Seoul on the 27th, various prospects regarding the future of AI and robots emerged. The panelists unanimously emphasized that 'the future cannot be determined', while presenting distinct challenges in their respective areas of expertise. Topics discussed included breakthroughs in manufacturing innovation, design principles of accessibility AI, conditions for the commercialization of home robots, and the Korean humanoid strategy.
The chairperson, Professor Yoon Sung-roh from the Department of Electrical and Computer Engineering at Seoul National University and former chairperson of the Presidential Fourth Industrial Revolution Committee, noted that "AI is a topic that simultaneously shakes the foundation of industry and society," adding that "the discussion today is a venue for seeking specific solutions while acknowledging uncertainty."
In the field of physical AI (AI that operates directly in physical environments), Ken Goldberg, a professor in industrial engineering at UC Berkeley and chairperson of the AI research lab, identified 'flexible material handling' as the biggest challenge. He explained, "Cables, fibers, and soft tissue are especially difficult for robots to handle, yet they are present in almost every product and living environment. For example, if robots could automatically recognize and untangle tangled cables, it would greatly help not just factory lines, but also places like concert halls, hospital operating rooms, and ships." He added, "The domain of handling '3D soft tissue,' such as cutting ingredients in a kitchen or suturing skin and tissue in surgeries, will be a new breakthrough in robot research."
Regarding the AI conditions for social minorities, Dona Sarkar, the head of AI & Copilot at Microsoft, emphasized the 'principle of participation.' Sarkar noted, "AI fails 100% if any stakeholder is missing at any stage of data collection, product development, or testing," citing the visually impaired support service 'Be My Eyes' as an example.
She explained, "In the collaborative process, general image data lacked the information that visually impaired persons actually need, such as expiration dates, medication labels, and signage." She added, "Therefore, we had to accumulate questions and logs left by users as new training data to create a separate model." She further remarked, "Avatars for sign language interpretation produced without the participation of deaf people have failed in over 20 countries," stressing that "disabled individuals must be directly involved in the design and validation processes to succeed."
David Qian, CEO of Ecovacs Robotics, addressed the realistic constraints of popularizing home robots. He said, "Homes are far more unpredictable spaces than factories," explaining that "when humidity rises in spring thaw, sensors may not operate correctly, and the conditions for motor operation can change simply with seasonal or material variations." He continued, "A universal robot that solves all household tasks will be a challenge that takes at least 5 to 10 years." He stated that "current specialized products that can reliably perform specific tasks like cleaning and laundry are more meaningful at this stage." He also pointed out, "If cars rely on long-range sensors, home robots must solve both short-range recognition and collision safety issues," adding that "using only commercial components is insufficient, so we are also developing our own sensors and motors."
Zhang Byoung-Tak, head of the AI research institute at Seoul National University and chairperson of the K-Humanoid Alliance, proposed strategies that Korea needs to prepare for in the AI era. He said, "While super-large AI faces disadvantages due to the data and capital gap, humanoids and physical AI are still at an early stage globally, presenting opportunities." He noted, "Korea has a complete ecosystem-based approach with demand companies, component manufacturers, batteries, AI semiconductors, and ICT infrastructure." He added, "We need to overcome the cultural and temporal differences between the manufacturing and AI industries," suggesting that "collaboration methods with the U.S. should focus on software and models, while with China, we should focus on manufacturing and parts."
During the Q&A session, 'AI investment bubbles' and 'developer jobs' emerged as key issues. Commissioner Goldberg remarked, "Rushed investments, like those during the dot-com bubble, can end in failure. However, just as the internet eventually proved its substance and gave birth to giant corporations, AI will also reveal its substance one day." He added, "The issue is timing, and we should be wary of those who make definitive claims about it." He further stated, "There will surely be increased demand for AI-based infrastructure like chips and data centers," likening it to the gold rush where those who sold tools like pickaxes and shovels made more money than those mining for gold, indicating that semiconductor and infrastructure providers will be the most reliable investment areas in AI proliferation.
Sarkar emphasized the uncertain job outlook for developers, saying, "No one can guarantee the situation a year from now." However, she noted, "As non-tech industries rush to adopt AI, developers are now required to understand the language of their respective industries and solve field-specific problems, rather than just coding." She remarked, "For example, to enter an airline, one must know the top 10 issues in the aviation industry, such as delays, lost baggage, and maintenance, and to join a financial firm, one must understand transaction risks and security requirements." She concluded that "individuals with both AI skills and industry understanding will be more competitive in the future."
Professor Yoon concluded the discussion by stating, "We have acquired the powerful tool of AI, but the key is how and where to utilize it," and he added, "The common conclusion shown in today's discussion is that we must recognize uncertainty and seek solutions through data and collaboration."