World Artificial Intelligence Conference Adopts 4 Papers from UNIST Professor Team

AI Graduate School Professors Sim Jaeyoung, Joo Kyungdon, Han Seungyeol, Kim Jisoo

AAAI Conference Held in Vancouver, Canada from 20th to 27th

Four research papers by UNIST faculty members have been accepted to AAAI (Association for Advancement of AI), a globally prestigious conference in the field of artificial intelligence.

These papers will be presented at the 38th Annual Conference, held for eight days starting from the 20th in Vancouver, Canada.

This year’s conference received over 12,000 submissions worldwide, the largest number ever. Only about 23%, or 2,342 papers, were accepted. UNIST’s research achievements came from four professor teams: Jaeyoung Shim, Kyungdon Joo, Seungryul Han, and Jisoo Kim.

Professor Jaeyoung Shim’s team developed computer vision technology that utilizes synthetic data for AI training. To enable AI to automatically identify specific individuals such as criminals or missing persons by analyzing video footage from CCTV and other sources, massive video data training is required. By using synthetic data, they drastically reduced the costs of data labeling and issues related to portrait rights infringement.

AI technology that enhances the vividness of virtual reality and games was also accepted. Professor Kyungdon Joo’s team introduces AI technology that precisely creates three-dimensional interactions between people in virtual reality.

In situations such as hugs and handshakes, virtual humans are automatically generated to match the other party’s movements, enabling natural interaction between virtual humans on a stereoscopic screen.

Professor Seungryul Han’s team presents multi-agent reinforcement learning technology. Multi-agent reinforcement learning is an essential AI foundational technology not only for games but also for autonomous driving and robotic swarm operations that require machine collaboration.

Professor Han’s team proposed a highly efficient algorithm that significantly reduces the number of cases that must be explored in existing multi-agent reinforcement learning.

Professor Jisoo Kim’s team developed a ‘bandit algorithm’ for online recommendations. This algorithm recommends similar items by considering users’ preferences in news, video viewing, shopping, and more. It features the ability to capture users’ diverse and complex preferences.

Established in 2020, the UNIST Graduate School of Artificial Intelligence has rapidly grown despite its short history by actively recruiting young and promising faculty members. Last year, a total of 31 papers were accepted at prestigious AI and computer vision conferences such as NeurIPS and CVPR, demonstrating its fast growth.