2022 AIGS Symposium, AI industry Gathers Together

The 2022 Artificial Intelligence (AI) Graduate School Symposium, hosted by the Ministry of Science and ICT, was held at the COEX Grand Ballroom for two days on August 18th and 19th. LG was the first private company to participate in this event as an organizer, together with the Artificial Intelligence Graduate Society (AiGS), the Artificial Intelligence Innovation Hub (AiiHub), and the Institute of Information & Communications Technology Planning & Evaluation (IITP).

The AIGS Symposium is a venue for government agencies, schools, and businesses to gather, exchange information, and develop a cooperative system for the advancement of the AI ecosystem in Korea. At this year's event, 16 AI graduate schools nationwide and 14 companies, ranging from large corporations such as LG, Samsung Electronics, KT, and POSCO to startups like Vuno, operated booths and participated in presentations and discussions.

On the first day of the event, CSAI Honglak Lee from LG AI Research delivered a keynote address titled “Learning compositional agents for real-world tasks.” CSAI Lee presented a method for AI agents to learn complex real-world tasks by combining and sequentially performing simple, low-level tasks. By learning through experience the interdependencies and correlations between subtasks, AI can promptly plan and execute tasks in new contexts. CSAI Lee stated, “By developing this, pre-learned knowledge can be expanded to new objects, or efficiently adapted to new tasks based on previous related tasks.” He added, “In addition, it is possible to generalize and improve learning efficiency by combining visual data with a large-scale AI model that has recently been in the limelight.”

The research being conducted in collaboration between AI Graduate School and LG was presented by Seunghwan Kim, leader of the Vision Lab at LG AI Research, under the title “Cases of Expert AI Research Collaboration Based on Multimodal Super-Giant AI.” In this presentation, Kim talked about an ongoing collaboration with the SNU-LG AI Research Center, which was co-founded by the LG AI Research and Seoul National University. According to Lab Leader Seunghwan Kim, the SNU-LG AI Research Center is developing a technology that recognizes atypical table structures by over-segmenting and merging cell location and connection information, as well as character recognition technology tailored to chemical documents using Latex-format OCR. Kim commented, “We are conducting research with Seoul National University on the EXAONE multimodal model for bidirectional image-text generation, and a research that predicts the entire body pose of AI human.”

AI Technologies Introduced by 8 LG Affiliates

LG's eight affiliates, including LG AI Research, LG Electronics, LG Display, LG Innotek, LG Energy Solution, LG Household & Health Care, LG U+, and LG CNS, participated in the LG booth, which was the largest this year's AI Graduate School Symposium Exhibition Hall. Among them, LG AI Research demonstrated the Super-Giant Multimodal AI EXAONE technology. It presented a vision model that converts the image to text and vice versa, as well as a language model that generates poetry and marketing phrases and summarizes the conversation of customer service and news articles. In particular, a poem written by EXAONE titled Advising Professor was well received by students.

 

LG Electronics brought a part of a machine to introduce a failure prediction solution that detects abnormalities in advance in driving components such as motor and cylinder using heat, vibration, and sound. Moreover, the AI vision inspection platform was used to directly demonstrate the process of inspecting the appearance and noise quality of household appliances such as washing machines, refrigerators, air conditioners, and dryers. While LG Display introduced a technology in which AI performs optimal display panel and circuit design on its own through reinforcement learning, LG Innotek unveiled AI technology that analyzes words, sentences, and contexts in patent documents in real-time to extract key patent information, such as problem-solving methods and fields of application.

 

LG Energy Solution introduced deep learning technology that predicts quality defects in advance during the battery activation process in a smart factory, and LG Household & Health Care introduced AI technology that recommends customized lipstick pigment formulation to cosmetic researchers within one minute. LG U+ exhibited an AI contact center. The presenter described callbots, which analyze and respond to conversations while interacting directly with customers, and advisors, which assist in consulting tasks such as listening to conversations and searching for relevant information while human agents interact with customers. While introducing deep learning technology for image recognition in industrial settings, LG CNS emphasized that it is conducting research on self-supervised learning to reduce labeling costs.

Who's the winner of the AI Graduate School Challenge?

 

LG AI Research and the Ministry of Science and ICT co-hosted the “AI Graduate School Challenge” to improve the practical skills of AI graduate students. AI Graduate School Challenge, which was held for the first time this year, is a competition in which graduate schools form teams and solve problems using real industrial data provided by companies. The award ceremony for the top five teams took place on the 18th. Since last July, 165 teams from 14 graduate schools have competed in a contest on the topic of “antigen/antibody response prediction for vaccine and immunotherapeutic development.” The winner went to the CVML team from Chung-Ang University. When applying for jobs at LG affiliates, the CVML team will receive benefits such as exemption from document screening and registration in the LG AI talent pool.

LG participated in the 2022 AI Graduate School Symposium for the first time as the organizer and actively participated in presentations and exhibitions. The event served as an opportunity to explore AI researches being conducted by other Korean companies and universities in addition to LG. We hope that Korea's capabilities in AI research will continue to expand through public-private partnerships.