Artificial intelligence (AI) is spurring innovation by combining and integrating with science and technology across various fields. Efforts to apply AI are also actively underway in the electromagnetic field. Accordingly, experts from the Korean Institute of Electromagnetic Engineering and Science gathered together to put their heads together to prepare for the future of electromagnetic waves and AI convergence.

The Korea Electromagnetic Engineering Society's Radio Education Research Group held a workshop on the topic of 'Fusion of AI and Electromagnetic Technology' at the Telecommunications Technology Association (TTA) in Seongnam City on the 31st.

In his opening remarks, Park Hak-byeong, chairman of the Radio Education Research Association, said, “In recent years, artificial intelligence has been introduced into various technological fields, and attempts are being made to utilize AI in electromagnetic wave design and analysis. I believe that electromagnetic wave design and construction methods will change significantly in the near future.”

This workshop featured lectures and discussions on the latest research trends in the convergence of AI and electromagnetic waves.

Through a total of four sessions, including △Physics AI & Generative AI (Moderator: Professor Kyung-young Jeong of Hanyang University), △Digital Twin (Moderator: Senior Researcher Hyung-cheol Moon of TTA), △Deep Learning and Electromagnetic Technology (Moderator: Professor Jae-young Jeong of Seoul National University of Science and Technology), and △Panel Discussion (Moderator: Hak-byeong Park of Samsung Electronics), insights on trends in AI technology application in the electromagnetic field and the latest research were shared.

Cho Chun-sik, president of the Korean Institute of Electromagnetic Engineering and Science (professor at Korea Aerospace University), said, “Recently, major companies are developing generative AI platforms to improve productivity, and the use of AI is expected to begin in earnest this year.” He also mentioned, “In the field of electromagnetic technology, a lot of research and workshops related to AI are being conducted, and deep learning technology is being widely applied in areas such as antennas, radar, medical care, noise analysis, and spectrum analysis.”

In addition, President Cho Chun-sik added, “I hope this workshop will serve as an opportunity for electromagnetic technology practitioners to consider the necessary capabilities and prepare for the future.”

■ A Dialogue on the Future of AI by Leading Electromagnetic Wave Expertsrong>

In the field of electromagnetic waves, a panel discussion was held to discuss various perspectives, including the convergence of AI technology and the resulting ethical issues, educational perspectives, future capabilities, and research directions.

When asked about the necessity of introducing AI in electromagnetic technology, Professor Kim Young-wook of Sogang University said, “Since electromagnetic wave (EM) simulation is a task that requires a long time, introducing AI can help in terms of optimization.” He also expressed his opinion, saying, “The problem of population decline in Korean society is directly related to the decrease in RF manpower, so in order to meet the ever-increasing hardware specifications, it will be necessary to reduce time, manpower, and cost with the help of AI.”

Professor Jeong Hae-jun of Hanyang University questioned the applicability of AI applications to electromagnetic waves in the direction of simply collecting and training a large amount of data, and pointed out that “it seems that the amount of investment for individual companies to create a foundation model equivalent to a large language model (LLM) in EM would be excessive, so we need to think about how to improve data efficiency.”

Additionally, concerns have been raised that individual companies are highly unlikely to disclose their data due to the constraints inherent in creating EM foundation models for PCB and RF design. Design data is a core element of each company, so it is difficult to create a model by exposing it externally.

Accordingly, Professor Kim Young-wook proposed a methodology in which each company individually models itself and then conducts federated learning based on those models to create a single foundation model.

Gathering data from each hospital is also a significant challenge in the development of hospital and healthcare applications. Therefore, federated learning, which trains AI models without directly sharing individual data, has been proposed as a way to protect individual companies' design IP while contributing to the development of EM foundation models.

Additionally, the panel's opinion raised concerns about the high cost of AI countermeasure design and the potential for challenges in applying advanced EM Foundation models to PCB design. From a practical perspective, the panel suggested leveraging these models in high-cost countermeasure designs requiring large layouts, such as system-level EMC.

■ Electromagnetic Wave-AI Convergence: Excessive Data and Resources Must Be Addressed First

Professor Jeong Hae-jun recently introduced research trends in which electromagnetic waves are used to implement artificial intelligence. He explained that AI operations consume excessive GPU resources, and that optical analog computing (OAC) with multiple layers can reduce the excessive power consumption in convolutional layers or hidden layers, and that the concept idea is being studied in related academic circles.

Additionally, a method for implementing edge detection of objects using metalens and metaelements was introduced. The extracted images suffer from image degradation, and research is underway to restore them using neural networks. This technology is expected to be applicable to radar image detection and radio wave technology.

Professor Jeong said that if the concept of deep learning backpropagation is introduced to electromagnetic waves and implemented with Maxwell's equations, it is expected to solve existing problems that require tens of thousands to millions of training data for simulation, and that he is conducting research on this.

Meanwhile, Professor Jeong's Radio-AI Laboratory is conducting research and development on: △ design techniques for radio/optical devices using AI algorithms; △ research and development of metamaterials, metasurfaces, and metalenses.