Professor Yeong-Jun Park, ‘From a memory superpower to an AI semiconductor superpower… the chance of success is ripe’ (Aju News, 2022.08.17)

2022-08-17l Hit 209

A contribution by Yeong-Jun Park, the chairman of the Intelligent Semiconductor Forum, and an emeritus professor at the Department of Electrical Engineering at SNU.

Yeong-Jun Park, the chairman of the Intelligent Semiconductor Forum

Recently, semiconductors have gone beyond a field that affects the national economy; it now affects national security. It has even come to a point where semiconductors are now considered an important war in the war between the US and China over supremacy. The reason is that products ranging from cars and robots to AI would not exist without semiconductors.

The AI field, which was initially thought of as a field limited to face recognition, is progressing to create voice samples, poetry, literature, and music, and it is now at the standard where making ‘digital humans’ is virtually possible. Such ‘digital humans’ that AI made based on dead people almost seem like living individuals.

We will soon be able to talk to Admiral Yi Sun-Sin. In order to teach an AI model the characteristics of Admiral Yi, it must be trained with a vast amount of data, so vast that the size surpasses one’s imagination. It is similar to a process where people form an image of Admiral Yi through extensive historical studies and then produce a film based on his story. It is hard to imagine the amount of data and calculations that will be needed to train the many digital humans of the future.

AI semiconductors are crucial in enabling endless development in such AI fields. The reason why Nvidia, a leading company in the field of AI semiconductors, could emerge is that the company is able to produce applications quickly and easily by carrying out massive calculations using AI semiconductor chips and software. However, the world complains that the products made by this company, which account for more than 70% of the global demand, are expensive, bulky, and consume a large amount of power.

Is there any way for Korea, the world’s number one memory superpower, to contribute to the solution to this problem? I would like to determine the eye of the hurricane here. This is because the numerous calculations and inferences performed by AI semiconductors have many similarities with the functions of memory semiconductors.

The ‘AI semiconductor industry growth support measures’ recently announced by the Korean government are in line with this context.

The idea is that innovative AI development can be achieved by utilizing the memory semiconductor structure cleverly instead of working based on existing structures. The strategy is to invest 1.2 trillion won in R&D and building a domestic IT infrastructure, build a cooperation system between large companies and SMEs, and train more than 7000 skilled experts in the field of semiconductors. Through this strategy, the government revealed its aspirations to develop the entire AI field and to develop into a system semiconductor design superpower.

Looking back, it has been 30 years since the government-led memory semiconductor research project began. At that time, the projects in which Samsung, Hyundai, LG Semiconductors, academia, and national research institutes participated greatly contributed to developing not only technology but also research personnel.

In comparison, the results of the non-memory field, that is, the system semiconductor R&D projects, were poor. Since then, the government presented the goal of raising the system semiconductor market share, which was about 3% at that time, by 10% every 10 years through the ‘System Semiconductor R&D Project’.

However, the projects always ended in failure, although there were results in equipment and material aspects. The new administration is trying to restart the system semiconductor industry. Learning from past experiences, what are the factors that affect the success of Korea’s semiconductor industry?

Firstly, Korea’s market semiconductor market share is increasing, and the overall research capacity has grown. The focused strategy that leverages Korea’s memory superiority to build AI semiconductors stands out. There is already a drive to actively participate in the industry, centering on memory conglomerates.

Secondly, the creation of applied fields. Compared to memory semiconductors, which are utilized everywhere, it is difficult to create demands for system semiconductors. From data centers that use semiconductors on a large scale to mobile devices, cars, and metaverses, major players have already dominated the demand. In addition, to change the existing system semiconductor, software has to change as well. The current government is well aware of this situation. They are setting up a strategy to jointly develop the semiconductor ecosystem by discovering demand sources such as data centers.

Thirdly, Korea’s semiconductor personnel are skilled, and infrastructure where talents and computers can participate on a large scale, is in place. The field of semiconductors has expanded to include experts in the field of computers, communication, even music, art, and psychology. Korea does have such skilled people in many areas other than the semiconductor field.

By making use of such success factors, I hope that Korea will be able to stand tall in the field of system semiconductors, which has always been deemed impossible.




Translated by: Do-Hyung Kim, English Editor of the Department of Electrical and Computer Engineering,