[Editor's Note] AI convergence is rapidly developing in the electromagnetic field. As AI technology begins to converge across all industries, discussions and research on the introduction of AI technology are actively underway in the EMI/EMC field. Accordingly, AI utilization and education for EM engineers and future technicians are expected to become important, and in this 'EMI/EMC Expert Talk', we met with experts on the topic of electromagnetic and AI convergence.

We recently met with EMxAI CEO Park Hak-byeong, who is preparing for a new beginning through EMxAI, and talked about the EMI/EMC outlook and AI convergence trends.

[Reporter] Please introduce yourself briefly.

[CEO Park Hak-byeong] I joined LG Electronics in 1992, and after working at Samsung Electronics, I have been developing electromagnetic wave (EMC, SI, PI) design and analysis methods required for home appliances and communication products for approximately 32 years. In October of this year, we founded and are currently operating EMxAI.

In addition, I will be serving as the Chairman of the Radio Education Research Group of the Electromagnetic Society of Korea in 2023, and I will be working with the society to improve the capabilities of electromagnetic field practitioners and graduate students.

[Reporter] You must be very active, I'm curious about your current status.

[Park Hak-byeong, CEO] This year seems to be an especially busy year with the activities of the Electromagnetic Society research group and preparations for starting a business. At the beginning of the year, we held the Workshop 2024 of the Korea Electromagnetic Society Radio Education Research Group with the theme of 'Electromagnetic Technology and AI Convergence'. After that, we received many requests for lectures, and recently, we finished the 24-year activities of the Radio Education Research Group by editing the Electromagnetic Technology Journal of the Electromagnetic Society.

Now, I have finished my long-term career as a researcher at a large corporation, established my own company, and am focusing on building it up.
[Reporter] I understand that you have been interested in and researching the convergence of electromagnetic waves and AI for quite some time. What is the trend of AI technology convergence in the EMC field?

[CEO Park Hak-byeong] Until the mid-1990s, EMC technology was developed by manufacturing actual products and verifying and improving them through experiments. Since then, it has been used for more than 20 years in a framework that utilizes methods such as simulation and knowledge-based design verification systems (Design Rule Check Systems).

AI technology has been introduced to electromagnetic technology for several years now, and recently, there has been a trend of expanding the introduction of various AI technologies, including large language models (LLMs).

Initially, AI such as CNN (Convolutional Neural Network) was mainly used in areas that analyze characteristics from images, such as radar and antenna radiation characteristics, and later reinforcement learning was used in various fields such as PCB optimal design, such as PCB power supply network design.

In particular, at the EMC/SI Symposium2024 held in Phoenix, USA in August of this year, academic papers using LLM were presented and lectures were given, attracting attention.

Such research is being actively conducted in research labs led by the EMCLab of the Missouri Science & Technology (MST) in the United States and Professor Erping Li's Lab at Zhejiang University in China.

[Reporter] With the generative AI boom, is the introduction of LLM becoming a hot topic in the electromagnetic field?

[CEO Park Hak-byeong] After the release of ChatGPT-3.5 in 2022, the emergence of LLM-based generative AI has become an important part of AI utilization in the field of electromagnetic technology. As in other fields, the scope of application of LLM-based AI in the field of electromagnetic technology is expanding, and its performance is also rapidly improving.

This opens the way for engineers unfamiliar with AI and computer languages to utilize AI in electromagnetic design and analysis.

Initially, it was difficult to design and analyze a simple patch antenna with chatGPT-3.5 through many trials and errors. Now, it is easy to design a 2X2 array antenna, simulate its characteristics, and report the results.

[Reporter] What is your outlook for the future?

[CEO Park Hak-byeong] We should pay more attention to the fact that AI technology can be easily utilized through LLM.

For example, let's assume a problem of automatically evaluating the electromagnetic characteristics of an array antenna as Pass or Fail using AI. In this case, it is necessary to train an existing AI model such as Alexnet into a model that can Pass or Fail the antenna, and this is called Transfer Learning.

It is difficult to implement if you do not have sufficient knowledge in the related field because it requires AI knowledge and programming skills, but it can be implemented even if you lack the knowledge through LLM.

LLM already has accumulated knowledge through learning about AI and computer languages, and by entering the data and simple text commands required for transfer learning, it can easily create code for transfer learning.

Google's CEO recently revealed that more than 25 percent of the programs the company produces are written using AI coding. It may feel like it is still in its initial stages and a long way off, but in the long run, it is expected to evolve into a form where AI will automatically perform complex electromagnetic wave design and analysis tasks when presented with drawings, specifications, and data, just as a chef would do the tedious cooking tasks when given a recipe and ingredients.

[Reporter] What is your outlook for the future of the EMI/EMC field?

[CEO Park Hak-byeong] With the trend toward product convergence and the expansion of wireless functions, digital noise suppression technology has become important to prevent radio frequency interference (RFI) between components within a product. In addition, with the trend toward enhanced portability, such as lower power consumption and miniaturization in mobile products, the importance of electrostatic discharge (ESD) countermeasure design technology has also increased.

The trend toward faster signal transmission speeds has made signal integrity (SI) technology an essential field along with EMC, and the range of products requiring the technology is expanding. Recently, with the emergence of new form factors such as foldable and flexible, the additional challenge has emerged of not only increasing the signal transmission speed, but also ensuring the same electromagnetic wave and signal characteristics in response to the folding and unfolding structure.

For example, in the case of foldable devices, the characteristics of the circuit (interconnect) through which signals pass are different when the product is folded and unfolded, which creates a complex challenge of having to maintain constant impedance and electromagnetic wave emission characteristics in all states.

In the future, electromagnetic technology is expected to continue to develop to reflect these product changes, and various AI technologies, such as physics-based neural networks (PINNs) that complement physics and neural networks, are expected to be actively introduced to complement the limitations of existing simulation technologies, such as analysis time, prediction errors, and complex utilization.
[Reporter] You recently started a business. Could you please introduce your company?

[CEO Park Hak-byeong] We founded the company under the name EMxAI. EMxAI stands for the fusion (x) of electromagnetic technology (EM) and artificial intelligence (AI).

Our main business areas include developing AI solutions that can efficiently implement electromagnetic technology and AI-based electromagnetic technology education that can expand and enhance the capabilities of circuit developers and electromagnetic engineers.

Through this, we aim to provide simulations for AI-based design and analysis in industries that require electromagnetic technology, such as electronics, communications, automobiles, and defense, and to contribute to improving the capabilities of related workers.

Currently, we are starting to develop educational courses in cooperation with educational institutions under the Radio Promotion Association of Korea (RAPA) and Telecommunications Technology Association of Korea (TTA), and we plan to provide related education starting in the second half of 2025. In the long term, we plan to develop AI solutions for electromagnetic wave design and analysis and provide electromagnetic wave design and analysis solutions using AI. We are preparing to link with existing design and measurement equipment.

[Reporter] Please tell us about your future activities and plans.

[CEO Park Hak-byeong] We plan to hold a workshop centered around the Electromagnetic Engineering Society to specify the direction of AI utilization in electromagnetic technology and expand the base of related technologies. We plan to continue our academic activities by participating in developing theoretical and practical training courses for graduate students and practitioners.

In addition, we plan to develop small-scale AI solutions starting in the second half of 2025 and release them through the company website. As the company is still in its early stages of establishment, we plan to focus on practical activities through the company rather than academic activities next year.

[Reporter] Lastly, please say a word to our readers.

[Park Hak-byeong, CEO] Our short-term goal is to get EMxAI on track. In 2025, we plan to develop systematic AI-based electromagnetic design and analysis education and provide it to practitioners and graduate students, and then develop AI solutions that can be used for electromagnetic design and analysis to kick off our activities in earnest.

Through this, we will strive to contribute to securing competitiveness in our electromagnetic wave industry. I hope that everyone working in electromagnetic wave technology will be happy and healthy in the coming year of 2025.

thank you