Following OpenAI's Generative AI ChatGPT, the shock wave of Chinese-made Dipsyck R1 is increasing anxiety over its future technology leadership
It is questionable whether Korea's industry will be able to survive the AI hegemony war with capital and manpower that are far short of those of the U.S. and China.
Yoo Hoi-joon, a leading semiconductor master scholar and chair professor of electrical and electronic engineering at the Korea Advanced Institute of Science and Technology (KAIST), will start the AI semiconductor CEO curriculum for the first time in Korea in March. The highest AI process combined with semiconductors is the first in Korea.
Professor Yoo's Contribution to the Field
Professor Yoo, who served as the chairman of the Semiconductor Engineering Association, heads the KAIST Graduate School of AI Semiconductor, the IT Convergence Research Institute, and the PIM Semiconductor Design Research Center. He gave the first keynote speech on AI semiconductors at the International Solid Circuit Society (ISSCC), a global semiconductor society, in 2019, was selected as the largest thesis in 2023, and he is the only Korean who participated as an operating member of the VLSI (ultra-dense integrated circuit) society dealing with semiconductor processes and circuits. Professor Yoo wrote a must-read "Design of DRAM" for semiconductor researchers and has studied AI semiconductors since 2008 and announced the world's first DNN accelerator in 2014.
Insights on Recent Developments
Regarding the recent DeepSeek shock, Professor Yoo said, "There was nothing new because the cost was significantly lowered by copying open AI with a known technology," adding, "Although there is not much technological difference from the AI algorithm Ex-One announced by LG Electronics' AI Research Institute last year, it has only attracted attention because of the huge market in China and will eventually remain 'Happening'."
Rather, he pointed out that attention should be paid to the fact that high-end Nvidia's AI chips, which have enjoyed monopoly status, are not sustainable this time. Roughly speaking, it was like cleaning a room with a tank.
![2.5 Fabrication](https://www.iue.tuwien.ac.at/phd/minixhofer/img79.png)
The Future of DeepSeek and AI Technology
It is also expected that DeepSeek will introduce optimized technologies using low-performance HBM (high-bandwidth memory), which will expand the low-end chip market. Among the learning and reasoning functions of real AI, high-spec chips are said to be essential for learning only.
Professor Yoo highlighted the trend of transitioning to the on-device AI market in the future, stating that Korean companies such as Samsung Electronics can have a clear competitive advantage in this shift. He emphasized the need for thorough responses from companies to adapt to the future changes in AI technology.
![Newsletter July 2024](https://s3-ap-south-1.amazonaws.com/ikarus3d-blog-prod/2024/07/unnamed-2.jpg)
Impact of Generative AI on the Industry
Generative AI is projected to be used like a cloud due to its high cost and high-performance computing power, but the era of on-device devices that use AI on devices like PCs (personal computers) is inevitable. They say that low-power DRAM (LPDDR) should be spread and a method of packaging it should be prepared.
Professor Yoo mentioned, "There are only three countries in the world that are playing AI semiconductors: the U.S., China, and Korea. "Europe and Japan are in disarray, and Taiwan only makes other people's," he said. "Korea is competitive in AI semiconductor applications. This is because the manufacturing base such as automobiles and home appliances is solid," he said with emphasis."
![Nvidia reaps benefits of AI growth](https://medias.yolegroup.com/uploads/2023/11/img-status-of-the-processor-industry_nvidia-revenue-q1-2021-to-q3-2024_nov2023-1024x662.jpg)
As AI Dae-hwan (AIX) is changing the framework of the industry, how well it uses AI has become the competitiveness of the industry. Individual companies should create solutions for AI by applying semiconductor hardware optimized for appropriate algorithms.
In particular, it plans to deal with cases such as bio, quantum computers, and robots as AI applications. This is a case in which the Coupang Logistics Center has controlled AI with an unmanned robot to reduce logistics costs by more than 40%, and pharmaceutical companies have shortened development time with AI.