SK Telecom (SKT) has set out to support the Industrial AI Transformation (AX) by applying its self-developed AI Foundation model to manufacturing sites for the first time. Starting with the steel and automotive parts industries, the goal is to digitize the on-site experience of skilled workers using AI and accelerate the response speed to process errors.

SKT announced on the 25th that it had signed a Memorandum of Understanding (MOU) with steel manufacturer KG Steel and automotive parts manufacturer KONEX, respectively, for field demonstrations of a manufacturing-specialized AI agent based on 'AX K1'.

SKT has completed the development of a demo version by collecting internal data, such as process error analysis reports, equipment manuals, and logs from the two companies, since last April.

AX K1 is a massive language model with 519 billion parameters, but only about 33 billion parameters are activated during inference.

SKT explained that it can reduce the computational burden on industrial sites by selectively operating only the necessary areas.

In the second half of the year, we plan to conduct field trials by deploying a demo version to the cold rolling line at KG Steel's Dangjin plant and the casting and processing processes at Konex, respectively.

Data collected during the demonstration process and field feedback are used to improve agent performance and are also planned to be used to train the subsequent model AX K2 currently under development.

The manufacturing industry is cited as a sector where AI adoption is slow due to the decentralized management of data across processes and departments and a high dependence on skilled workers.

It is pointed out that if the 'knowledge silo' phenomenon, where core know-how is concentrated among specific workers, intensifies, on-site competitiveness could weaken if those personnel leave.

In addition, there are many environments where external cloud integration is difficult due to process security requirements.

SKT announced that the AX K1 supports a closed on-premise environment along with the cloud, enabling operation without exporting data externally.

SKT It was added that it had signed an MOU with the Ministry of National Defense last month regarding the utilization of its proprietary AI Foundation model, and plans to expand the scope of application to finance, the public sector, and healthcare in the future.

Jeong Seok-geun, Head of SKT’s AI CIC, stated, “For manufacturing sites where security is critical, our proprietary AI Foundation model, which allows for data utilization without being leaked externally, is an effective solution.” He added, “Starting with our collaboration with KG Steel and KONEX, we will accelerate the AI transformation of the manufacturing industry and expand the application cases of our proprietary AI Foundation model.”