The automobile industry is also required to innovate in order to respond to paradigm shifts that go beyond the advancement of AI technology. Industry insiders gathered together to discuss strategies for responding to AI technology being integrated into mobility.

On the 18th, the Korea Automobile Research Institute held the 18th Asset Management Report event at the COEX Startup Branch in Samseong-dong, Seoul.

This event, held under the theme of ‘AI and Mobility Innovation, Designing Smarter Mobility,’ was attended by approximately 100 people including officials from related organizations and companies related to AI technology and mobility, including the Ministry of Trade, Industry and Energy, Gwangju Institute of Science and Technology, Artificial Intelligence Industry Convergence Business Division, LG Electronics, AiMo, and Applied Intuition.

Starting with the welcoming speech by President Jin Jong-wook of Hanjayon, the event included: △Mobility+AI Response Strategy for the Ultra-Gap Era (Kim Jong-won, Dean of Gwangju Institute of Science and Technology AI Graduate School), △Trends in Mobility Application Cases Utilizing Ultra-Large-Scale Generative AI Technology (Kang Byeong-su, Head of Big Data and SW Technology Division, Hanjayon), △Introduction of AI Trends and AI Industry Convergence Complex Infrastructure (Kwak Jae-do, Head of AI Industry Convergence Business Unit), and △Presentation of Leading Companies.

Jin Jong-wook, CEO of Hanjayeon, said, “AI is a hot topic and is leading new changes related to mobility,” and added, “It is expected to bring about new changes in all areas including design, production, operation, and service.”

Meanwhile, the Gwangju Regional Headquarters of Hanjayeon and the research results and technology consulting booths of the big data and SW technology sectors, as well as promotional booths for five innovative companies including Applied Intuition, Bitsensing, Mobizen, Vessel AI, and Data Alliance, were set up.

Kwak Jae-do, head of the Artificial Intelligence Industry Convergence Business Group, said, “Unlike in the past when companies unanimously insisted on cost support, these days there are many cases where they say, ‘Well-processed, high-quality data is needed’ and ‘There is a shortage of skilled AI personnel,’” emphasizing the importance of securing data.It has been confirmed that this is happening.

■ Keynote speech “AI, competing at the level of an arms race”
Kim Jong-won, Dean of Gwangju Institute of Science and Technology's AI Graduate School, gave a keynote speech on the topic of mobility and AI response strategies in the era of hyper-gap.

“Data centers do not define all infrastructure,” said President Kim. “They must be supported by the maturity of society-wide data and computer infrastructure.”

Data centers play a symbolic role in AI infrastructure and industry, and when AI is combined with mobility, AI can only be powerful when there is sufficient data.

President Kim said, “Mobility must move beyond the range of vehicles that move, that is, the ‘city,’” and urged that the construction of digital twins and AI infrastructure be viewed broadly as a smart city phase.

■ Synthetic data for autonomous driving, must overcome Habsburg AI
Kang Byeong-su, head researcher at Hanjayeon Big Data and SW technology division, is using ultra-large-scale generative AI technology for mobility. Introduced application case trends.

Researcher Kang explained, “AI development is divided into model-centered and data-centered, and is moving more toward data-centered development.” Complementary development is expected in the future, but at the current stage, methods are needed to train physical AI with a small amount of data.

Researcher Kang pointed out that while using generated data to learn with less data, we must be careful of the 'Hapsburg AI phenomenon'. Hapsburg AI is a problem in which performance deteriorates when continuously learning generated data, causing the data distribution to become increasingly lopsided and the performance itself to drop.

Researcher Kang said, “We are faced with the task of continuously injecting high-quality real data,” and suggested, “It is important to create high-quality data for autonomous driving generation data as well. We need to be able to create a feedback structure that can automatically verify and evaluate by implementing an evaluation model.”