Prof. Kyoung Mu Lee, “In the U.S., ‘top talents’ in AI are paid ten times the highest-paid professors at SNU” (The Hankook-Ilbo, 20191226)

2020-03-02l Hit 42

'AI expert’ SNU ECE Professor Kyoung Mu Lee
Korea announces aim to globally rank 3rd after the U.S. and China by 2030 according to AI national strategies
In MIT, a fundraising of 1.1 trillion won to found an AI college… SNU has only a few tens of billion won

In an interview on December 23rd with the editorial member Jeong Yeongo (정영오), Professor Kyoung Mu Lee (left) said that although there is worldwide competition for the lead in AI technology, Korea can catch up quickly with its advantage in semiconductor technology and high-speed information network.

On December 17th, the government announced the ‘AI National Strategies’. In a promise to the Korean people, it stated that with AI technology, it would generate an economic value of 455 trillion won and rank 10th in quality of life by year 2030. I met SNU ECE Professor Kyoung Mu Lee, who helped develop the national strategies as a Korean expert in the AI field, for his opinion on the current strengths of Korea’s AI and the plausibility of the AI national strategies.

-Fervor surrounding AI has heightened, leading to the government announcement of ‘AI national strategies’. For the general public, however, the sentiments attached to AI concern anxiety that outweighs anticipation. The ‘national strategies’ also predicts that 10% of Korea’s jobs will be replaced with AI.

“AI technology itself is on neutral grounds. It contains the features of both utopia and dystopia. Based on how it is applied, it could either boost our economy or grow into a monstrosity that seizes jobs and impoverishes the general public. With every major innovation in technology such as the Industrial Revolution, there was significant impact upon society. Eventually, however, there were more jobs. A future led by the AI will also follow such a trail. AI has already settled into our surroundings. People are now familiar with using AI speakers that they order vocally or features for automatic translation of foreign languages. AI is being actively applied in fields of expertise. For example, it is used for analysis of investment risks in the financial sector and diagnosis in medicine.”

-The impact of AI will leave out no field. Even the arts as well as the medical field and the judicial system will be affected. What is the safest workplace in ten years?

“The management consulting firm McKinsey foresees that by 2030, 27% of domestic labor will be automized. One prediction states that in 120 years, all jobs over the world will be automatized. In Korea, by 2030, 7 million jobs will disappear and, in its stead, 7.3 million will be created. There will be 300 thousand more jobs. Of course, since the form of jobs will change, it is important that people are trained again for the jobs. How society adapts to the advances in AI becomes a matter of social and political issues rather than technology. As our human resource pool is limited, means of social and policy-wise support for people to flexibly adapt to changes should be prepared and implemented. I believe that even with AI, it will comparatively take longer to replace jobs that require expertise related to the human emotion such as culture, the arts, and psychotherapy.”

-Perhaps one has no choice but to choose a job that one wants rather than be concerned over competition with AI.

“A job is both a means of self-realization and earning your daily bread. The importance of occupations in the latter case will dwindle in the future. Eventually, the issue of how to distribute various goods and profits produced with AI will rise in importance for the social management. A society where humans leave the work to sustain life to AI and select professions for self-fulfillment is no longer a fantastical notion.”

-To create such a society, the ‘AI national strategies’ announced by the government must be realized first. The aim is to rank 3rd in the digital field and generate an economic worth of 455 trillion won by 2030. Could you elaborate on the economic value of 455 trillion won?

“The estimate was calculated using McKinsey’s analytical methods. Based on the previous outlook on economic growth, the effects of sales increase, cost reductions, and improved consumer welfare were estimated. Calculations showed that the additionally generated wealth will amount to at least 225 trillion won and at most 455 trillion won. As McKinsey conjectured that there will be an economic effect of 7 quadrillion won generated over the globe in 2030, approximately 6.5% is to account for Korea. Since Korea’s portion in the worldwide GDP is currently about 1.9%, you can say we are being ambitious. However, if we realize the goal to rank third after the U.S. and China in the digital field by year 2030, such an economic effect will be plausible.”

-Will Korea’s strength in AI be led by hardware as is the case for semiconductors? The national strategies explicitly state the aim to rank first in AI semiconductors.

“Both software and hardware are important. Eventually, we must co-design the two. However, Korea has to set a strategy where we choose and concentrate based on our given strengths. AI semiconductors are semiconductors that apply AI techniques by concurrently processing memory and calculations, similar to the human brain. Novel semiconductors that merge memory with microprocessors are called PIM (Processing In Memory) semiconductors. PIM semiconductors allow the mimicry of the human brain by permitting the transition from the conventional CPU-centered computing to memory-centered computing. The strategy is to become the champion in AI semiconductors based on our world-leading memory semiconductor technology. To realize this plan, our globally top-ranking information network infrastructure will help expedite the application of AI technology. Currently, the U.S. and China has taken the lead in investment and research for AI semiconductors. However, Korea, with its memory-related tech and manufacturing process, will be able to catch up with them in the near future.”

-In the end, advances in AI technology are driven by people. In reality, however, the SNU Graduate School of Data Science, to be open starting next year, has yet to fill its faculty quota due to the low salaries offered. With even SNU struggling to find faculty members, would raising talents work out as planned?

“In the U.S., top talents are usually rewarded more than ten times the highest-paid SNU professors. Funding for MIT’s AI college at the time of its founding amounted to 1.1 trillion won. Donations for SNU’s AI research and education add up to a few tens of billion won. The difficulty in recruiting top level talents with such wages and research environments is evident. On a bright note, with such considerations in mind, the government stated that it will relax regulations and allow professors of AI related departments to hold other offices in companies starting next year. Both the U.S. and China permit concurrently holding offices in companies to gather top level talents in AI as well as handle their high pay. Recruiting top talents from the outside is not enough to supply the necessary human resource pool. It is vital that we deal with the question of how to raise the quality of domestic AI talents. Graduate students in AI should be allowed to concentrate solely on their research, but things are tougher in reality. For instance, there is a wide gap between Korea and the U.S. or China in the quality of equipment necessary for research. There needs to be bolder investments and government support to increase the number of AI majors each year and to produce ten thousand experts in the AI field annually from 2030.”

SNU ECE Professor Kyoung Mu Lee

-More students are applying for AI related departments. What qualities do you have to attain to become an expert?

“From what I know, the government’s goal to raise ten thousand talents in the AI field is a two-track plan. The plan is to acquire two thousand top professionals and eight thousand people specializing in the field. The top professionals are key AI talents who have completed related majors and have their Master and Ph.D. degree. They will be raised through AI graduate schools, BK21+ projects, and priority research centers. The other eight thousand have Bachelor’s degrees and are trained through SW centered universities, founded or expanded AI related departments, and related academies. This track can be replenished with various people including those from the humanities. In particular, it would be nice if there are applicants with the creativity and passion to merge AI and the fields in which they have experience.”

-For advances in AI, data accumulation and application are as important as human resources. Strength in this area is determined by the effectiveness of pertinent regulations. What is the case with Korea?

“When you look into AI from an industrial point of view, there are several success factors that are noted. The four factors of human resource, data, policy, and market must be well managed. The U.S. and China are far in the lead in those four criteria. Korea is lagging behind in all of those categories. We have a limited source generation of data, and there are regulations that pose obstacles for data access. The ‘three acts concerning data’ which are pending in the National Assembly must be enacted soon as the first step to a solution. The task of constructing a management system where established data can easily be shared depends on policies. In particular, there is an urgent need for a mutual sharing and managing system for public and private data. A national organization that focuses on this task should be established, thus starting the gathering, producing, connecting, and combining of public and private data. China, due to the characteristics of its system, is at the head of this field. The government is actively intervening, designating fields for private businesses such as autonomous driving for Baidu and healthcare for Tencent. On the other hand, the U.S. is entrusting data compilation to the private sector.”

-The government aims to rank third after the U.S. and China in the AI field by year 2030. How wide is the gap currently? Is it too late?

“Although we are late, I think it is an achievable goal. You could have doubts if you only look toward the U.S. and China, but when you compare Korea with other competitors, we are acquiring strength quickly. According to a research by China’s Tencent in 2017, there were 22 thousand key AI talents in the world. The U.S. had half. Korea’s portion was less than one percent. Korea was evaluated to have less than 200 experts. In October this year, ‘International Conference on Computer Vision(ICCV) 2019’, the top conference in AI, was held in Seoul. I was there as the General Chair. It was a conference where the most actively researching AI specialists in the world present their research results and major AI related companies participate to display the fruit of their developments. This year, 1075 papers were published in this conference and China surpassed the U.S. with 362 papers to the latter’s 313. Korea had 44 publications. In two years, the number of key talents had increased to add up to more than 10% of China or the U.S. In particular, the fact that as a nation, Korea ranked fourth in the number of papers published was an auspicious sign. The key to achieving the goal will depend on the number of domestic AI talents worthy of worldwide recognition.”

SNU ECE Professor Kyoung Mu Lee…

  • received his Master’s degree in Control and Instrumentation Engineering at SNU and his Ph.D. degree at USC, the U.S. He is serving as the Associate Editor in Chief for IEEE TPAMI (Transactions on Pattern Analysis and Machine Intelligence), a globally renowned AI journal, and President of Korean Conference on Computer Vision. In October, he successfully oversaw the management of the ‘International Conference on Computer Vision(ICCV) 2019’, a world renowned conference in AI, as the general chair.

Translated by: Jee Hyun Lee, English Editor of Department of Electrical and Computer Engineering,