Robots for Dangerous Tasks: On-Site Discussions Lag

While the idea that "robots should do dangerous jobs instead of humans" is frequently mentioned in discussions about the spread of physical AI, related discussions at actual hazardous sites are progressing slowly compared to the need. In manufacturing sites with a high concentration of high-risk tasks, there's consensus on the necessity of adopting physical AI, yet implementation is sluggish. Meanwhile, in repetitive task environments where automation is already advanced, labor unions are showing vigilance. The reaction from industrial sites to physical AI varies depending on the nature of the work rather than the industry itself. Industries with a high proportion of repetitive tasks show relatively greater concern about changes in employment and working conditions, whereas industries with a high proportion of dangerous tasks express higher expectations for improved safety.

The automotive industry is considered a prime example of manufacturing with a high degree of automation. The Korean Metal Workers' Union has included a demand for prior impact assessments with the union before AI adoption in its negotiation proposals this year. The union insists that a comprehensive review is needed for overall impacts, including job losses, wages, work methods, labor intensity, and safety issues. While not unconditionally opposing physical AI, the union believes it should be adopted only after its effects on workers are assessed beforehand and in a manner that complements human labor. It's analyzed that the relatively high potential for physical AI to replace existing labor in a production structure based on repetitive tasks on conveyor belts leads to the union's vigilance.

In contrast, the atmosphere in the steel industry is somewhat different. The POSCO Labor Union has clearly stated its position of not opposing the adoption of physical AI, viewing proactive technological implementation as necessary for ensuring safety, given the numerous hazardous processes involving high temperatures and heavy materials. Union representatives stated that steel plants have many dangerous sites and AI should be adopted swiftly to allow workers to work more safely. However, they added that the process of discussion between labor and management, reflecting on employment issues and on-site voices, is crucial, rather than unilateral adoption. Indeed, POSCO is expanding automation in some processes, such as testing autonomous driving technology for torpedo cars that transport molten metal.

The shipbuilding industry also agrees on the need for physical AI adoption for safety reasons, but explanations suggest the feasibility of actual implementation differs from the automotive industry. The ship construction process involves complex curved structures, confined spaces, and overhead work, making standardized automation difficult. Tasks like welding and painting still require significant human intervention, and even with technological advancements, it's unlikely that the entire site will be replaced by robots within a decade. The shipbuilding sector faces a deepening labor shortage, with a high proportion of foreign and subcontracted workers. It is pointed out that the structure of utilizing relatively cheaper labor persists over adopting expensive physical AI. A representative from the Samsung Heavy Industries union echoed the sentiment that robots should indeed perform dangerous tasks, but noted the practical implementation challenges differ from the automotive industry.

The labor sector as a whole does not oppose physical AI itself, stating that physical AI taking over high-risk and long-hour tasks performed by humans would benefit workers. The key issue is not the technology adoption itself, but how to resolve the employment, redeployment, and training issues for existing workers. They added that labor and management must consult beforehand to coordinate manpower allocation and training, and if necessary, the government should also engage in discussions on creating new industries and jobs.

Professor Kim Sung-hee from the Korea University Labor Problems Research Institute explained that while automotive assembly lines, with their high repetitiveness, offer significant potential for physical AI replacement, shipbuilding and steelmaking, due to their non-standardized work characteristics, are closer to a complementary relationship rather than a replacement. He added that for hazardous industries, physical AI is expensive and prone to rapid damage in extreme environments, and there's little incentive for adoption when low-wage labor remains cheaper. While the market talks about applying physical AI directly to the field, he explained that considerable time could pass before actual implementation. Experts analyze that the differing temperatures surrounding physical AI adoption across industries stem from work characteristics and replacement potential, rather than the assertiveness of labor unions or wage levels.