SK Telecom will apply its self-developed proprietary artificial intelligence (AI) foundation model to steel and auto parts manufacturing sites. The plan aims to solve the "Knowledge Silo" problem, long cited as an obstacle to AI adoption in industrial settings, and to accelerate the pace of AI transformation in manufacturing.
SK Telecom said Tuesday that it had signed strategic memorandums of understanding (MOU) with steelmaker KG Steel and auto parts manufacturer Connec, respectively, to pursue field validation of AI agents based on its proprietary AI foundation model.
Since April, SK Telecom has obtained manufacturing site data held by KG Steel and Connec, including past process error and accident analysis reports, equipment manuals, and facility logs. Based on this, the company developed a demo version of a "manufacturing-specialized AI agent" built on its proprietary AI foundation model "A.X K1."
A.X K1 is a hyperscale language model with 519 billion parameters. It has strengths in handling complex commands and tasks, and only about 33 billion parameters are activated during the inference process. The company explained that while the overall model scale is large, only the necessary portion operates at the actual usage stage, making it relatively efficient to use even in industrial settings.
SK Telecom, KG Steel, and Connec will conduct field validation by applying the demo version of the manufacturing-specialized AI agent based on A.X K1 to actual processes in the second half of this year. KG Steel plans to deploy the AI agent on the cold rolling line of its Dangjin plant, which produces coated steel sheets, while Connec will deploy it in casting and processing operations. KG Steel and Connec will provide SK Telecom with more manufacturing process data in real time, and SK Telecom plans to enhance the performance and inference speed of the manufacturing-specialized AI agent based on this data and field feedback obtained during the validation process. The company will also pursue feature expansion.
The manufacturing site data obtained during the validation process will also be used to train the follow-up model A.X K2, which is currently under development. SK Telecom plans to review the commercialization and actual adoption of the manufacturing-specialized AI agent after the validation is completed. If necessary, it will also consider replacing the model with a follow-up series of SK Telecom's proprietary AI foundation model.
Manufacturing has been cited as a field where AI adoption is not easy. This is because the digital transformation of field data has been slow, and accumulated data is often generated and managed separately by process and department, making it difficult for AI to use it in an integrated manner. The fact that work processing methods vary according to workers' skill levels and experience has also been pointed out as a limitation of AI adoption in manufacturing. In particular, the "Knowledge Silo" phenomenon, in which core know-how remains only with specific skilled workers, is cited as a representative challenge at manufacturing sites. This is because when veteran workers retire or change jobs, the field experience and problem-solving methods accumulated over a long period may disappear with them.
SK Telecom plans to convert data scattered throughout processes and the experiential knowledge of skilled workers into digital assets, and introduce AI agents that have learned from these to manufacturing sites. Through this, the company expects to identify the cause more quickly when errors occur during processes and reduce response time, thereby improving production efficiency.
Meanwhile, SK Telecom is also expanding the application fields of its proprietary AI foundation model. Last month, it signed an MOU with the Ministry of National Defense for the "Use of the Ministry of Science and ICT's Proprietary AI Foundation Model in the Defense Sector." Following its decision to apply the proprietary AI model to the defense field, where security and data sovereignty are important, the company has now expanded the scope of application to steel and auto parts manufacturing sites.
"In manufacturing sites where security is important, a proprietary AI foundation model that can be utilized without taking data outside can be an effective solution," said Jung Suk-keun, head of SK Telecom's AI CIC. "Starting with our cooperation with KG Steel and Connec, we will accelerate the AI transformation of manufacturing and continue to expand application cases for our proprietary AI foundation model."
