LG AI Research Holds 『LG AI Talk Concert - The Best of 2021』, Sharing Achievements in Light of its First Anniversary

The “LG AI Talk Concert – The Best of 2021” will be held on December 14th with the purpose of sharing the achievements made by LG AI Research throughout the year. The LG AI Talk Concert is a seminar-type event that introduces the latest AI technologies quarterly, however, the last LG AI Talk Concert of each year is held as an external event in order to present the research results of LG AI Research.

The “LG AI Talk Concert – The Best of 2021” begins with a keynote speech from Kyunghoon Bae, the representative of LG AI Research, summarizing its 2021 research results. During the keynote, the results of the LG AI research thesis selection from major overseas conferences in 2021, the progress status on a number of projects carried out with LG Group affiliates, as well as the results of the super-giant AI research will all be presented. Being the first time in unveiling said research results of LG AI Research, which declared its development of the world’s best super-giant AI in May, its keynote is receiving the global spotlight.

In three detailed sessions following the keynote, various members will present the contents of specific researches. A distinctive feature of this LG AI Talk Concert is that it allows you to hear the vivid voices of the members who directly conducted the research, breaking away from the usual composition where all announcements are made by the head of the organization. In addition to its official achievements, the 12 members plan to convey a variety of interesting topics including the background of the researches as well as episodes that pertain to the research process.

LG AI Talk Concert Program

The first session is one that regards the Fundamental session, introducing the latest research papers of LG AI Research. Moontae Lee, a Lab leader of the Fundamental Research Lab, will be taking the lead as chairperson. As the first presentation of the session, Sungryull Sohn will be presenting on the methodology of efficient learning in an environment where rewards are sparsely given in relation to reinforcement learning, a subject that is receiving great attention in academia. The relevant research paper (Shortest-Path Constrained Reinforcement Learning for Sparse Reward Tasks) was selected at the prestigious International Conference on Machine Learning (ICML) 2021.
Next, Hyeongrae Ihm will introduce the algorithm for “composing artificial intelligence.” His paper on composition AI, which creates a natural ensemble of multiple instruments, was selected at the NeurIPS 2021, another prestigious international conference (Composer AI with tap-to-pitch generator: AI-based Music Composition). Lastly, Yeonsik Jo from Vision Lab will introduce the paper “Rethinking Deep Image Prior for Denoising” (selected by the ICCV 2021), which utilizes the features that occur in the learning of the Convolutional Neural Network (CNN) for noise removal.

The second session is an Applied AI session that introduces the achievements of a number of artificial intelligence projects that LG AI Research has been working on with various affiliates of the group. Woohyung Lim, a Lab leader of the Data Intelligence Lab, will lead the session as the chairperson. Firstly, Sehee Han, a leader of the Materials Informatics(MI) Squad, will present the progress and performance of the Materials AI research that discovers new materials such as bio-drugs, battery materials, and luminescent materials.
Next, Squad Leader Byungjun Kang from Vision Lab will introduce the progress and performance of the next-generation vision inspection research on unsupervised learning methods, that have surpassed the performance of existing supervised learning methods, and Squad Leader Yeseul Sim from the Data Intelligence Lab will introduce AI-based battery abnormality detections. The Data Intelligence Lab Squad Leader Jinseok Yang will present on projects related to deep learning-based demand forecasting that has improved the efficiency of forecasting that is done wid minute amounts of data, and Squad Leader Kyungkoo Min from the Language Lab will present on the progress and performance of the Machine Reading Comprehension (MRC) research which improves the document comprehension ability of artificial intelligence. Since all of these projects far surpassed the performance of the existing methods by applying the latest AI algorithms to the industrial sites, it will provide good references as being the Best Practices in related academia as well as the industry that performs the AIX application tasks.

The last session, with CSAI (Chief Scientist of AI ) HongLak Lee participating as the chairperson, will be a session on the super-giant AI that is expected to receive the most amount of attention at the “LG AI Talk Concert – The Best of 2021.” While Squad Leader Sihaeng Lee, Taehoon Kim, Janghoon Han, and Hyunjik Jo, who all will also be participating at the super-giant AI project, are planned to make presentations, the title and contents of said presentations will be private until just before the keynote. However, the scope and level of the presentation is expected to be significantly different from other super-giant AI related cases that have been disclosed so far, and it will portray LG’s philosophy of looking at super-giant AI beyond simply the level of research performance disclosure.

“LG AI Talk Concert – The Best of 2021,” where you can check the current status of the latest AI research conducted by LG AI Research, will be released through the LG AI Research YouTube channel starting from 9am on December 14 and will also be streamed live.

LG AI Research Youtube (Link)