'Physical AI: From a Performance Race to a Field Deployment Race' Report

There are calls for Korean companies to rapidly turn high-difficulty manufacturing processes into data assets, given their weak model competitiveness in the Physical AI field.


Samil PwC announced Wednesday that it had published a report titled "Physical AI: From a Performance Race to a Field Deployment Race—Field Deployment Accumulates Data, and Data Widens the Technology Gap." The report analyzes the Physical AI ecosystem, examines domestic and overseas market conditions and bottlenecks to wider adoption, and presents recommendations for suppliers and adopters.


The report found that although Korea has strong semiconductor memory capabilities and world-class manufacturing facilities in industries such as automotive, electronics, and shipbuilding, its competitiveness in actuation components and general-purpose artificial intelligence (AI) models is relatively weak. It said that, rather than competing with U.S. models and infrastructure or China's low-cost hardware, Korea needs to turn its manufacturing processes in sectors such as semiconductors, shipbuilding, batteries, and automobiles into data assets that can be used to train Physical AI.

Samil PwC: "Korea's Physical AI Sector Must Turn High-Complexity Manufacturing Processes into Data Assets" View original image

The report recommended that Physical AI suppliers focus not on competing with general-purpose models or finished products, but on AI models specialized for specific industries, scarce components such as precision robotic hands and tactile sensors, robot-as-a-service (RaaS) models, and capabilities in field operations and systems integration. It also stressed the need to secure ownership of and usage rights to data generated during robot deployment.


It said that adopters should take on the role of providing key data. The report recommended starting with processes where results are easy to verify, such as repetitive tasks, and clearly defining ownership of and access to operational data, as well as whether the data may be used for model training or shared with third parties, when signing contracts with robot suppliers.



"Competition in Physical AI is expanding beyond the performance of individual models and robots to encompass how reliably they can operate in real-world settings and how continuously they can be improved," said Kim Jeongyeon, a partner at PwC Consulting. "Korean companies should leverage high-difficulty manufacturing sites as strategic assets to accumulate real-world data and build a virtuous cycle that connects this data to training and performance improvements."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing