Kim Jae-joon, a professor in the Department of Computer Engineering at Seoul National University, has been selected as the recipient of the Science and Technology Person of the Month award for May. The Science and Technology Person of the Month award is given to one researcher each month who has contributed to the advancement of science and technology through excellent research and development achievements, along with a ministerial award from the Ministry of Science and ICT and a cash prize of 10 million won.
Professor Kim was recognized for developing lightweight artificial intelligence (AI) models and semiconductors that efficiently compute these lightweight models. A semiconductor accelerator refers to a semiconductor chip optimized for calculations in specific applications rather than general-purpose computing.
Recently, various AI models, including large language models (LLMs) like ChatGPT, have rapidly spread across industries and daily life, consuming enormous computing resources. This sharp increase in power usage has emerged as a social issue.
In response to this issue, research aimed at improving the efficiency of AI models is actively underway. Representative technologies include software-centric lightweight techniques that compress the size of AI models and hardware-centric semiconductor accelerator technologies that efficiently compute the lightweight models.
Professor Kim has developed a new lightweight model by integrating both approaches while also creating dedicated semiconductor accelerators that efficiently support the lightweight models. His research achievements were published in the June 2022 issue of the International Journal of Solid-State Circuits (JSSC), and results related to low-power AI semiconductors were presented at the International Conference on Learning Representations (ICLR) in May 2023.
Professor Kim noted, "I want to continue to focus on research in low-power integrated circuit design that can minimize power consumption, contributing to the sustainable development of AI," adding, "I expect that this research result will be utilized as a key foundational technology that enables AI computation in mobile environments where power consumption is critical."