LG AI Research Attends Summer Conference by the Korean Artificial Intelligence Association

LG AI Research participated in the 2021 Summer Conference hosted by the Korean Artificial Intelligence Association.

The conference, which was held online, from July 8 to 10, was attended by Professor Sergey Levine of UC Berkeley, one of the world’s most authoritative researchers in reinforcement learning and optimization, and Professor Stephen Boyd of Stanford University as plenary speakers and Jaeseung Jeong of KAIST and Sangkyun Cha of Seoul National University as guest speakers. Additionally, there were other various programs such as 5 tutorial lectures, 7 planning sessions, 2 special sessions, and a research paper session.




LG AI Research prepared 'LG AI Research Special Session' on July 9 and introduced its latest research and LG’s Super-giant AI research directions. During the presentations regarding the latest research, Researcher Sungha Choi from Vision Lab presented RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening and Researcher Janghoon Han from Language Lab gave a lecture on Fine-grained Post-training for Improving Retrieval-based Dialogue Systems.

RobustNet presented by Choi is expected to be applicable to autonomous driving in the future with vision technology research, which identifies the areas that cannot be recognized traditionally, with a new method called selective whitening loss. 

Han explained his research which used fine-grained post-training to solve the problem of a language model being unable to directly identify relationships between speeches in the context when applied to the retrieval-based dialog system.

RobustNet presented by Choi (left) and fine-grained post-training explained by Han (right)

Following presentations on the latest research, lab leader Jungkyu Choi of Language Lab, LG AI Research introduced LG’s Super-giant AI research directions. Lab leader Choi set forward LG AI Research’s Super-giant AI directions—namely integrated expert AI, challenging the impossible AI, and imagining and creating AI—and presented specific plans for these said directions. LG AI Research declared the development of Super-giant AI in May this year and announced its vision to obtain one of the world’s top 3 Super-giant AI (link: LG AI Research Rises to the Challenge of Super-giant AI).

This Summer Conference by the Korean Artificial Intelligence Association is the second conference held by the association since its foundation in 2016 and is prepared to contribute to advances in AI in Korea as well as serving as a platform for technology exchanges. LG AI Research, Korea Midland Power, Naver, LG Innotek, KT, VUNO, ETRI, and KAIST all sponsored the conference.