LG

■ Improved the accuracy of the “Explainable AI” technology that allows humans to easily understand the reasoning behind AI's judgements
 □ Extensive applications in fields where reliability is crucial, such as medicine, finance, and law 
 
■ Developed AI technology that reduces memory usage by 40% while maintaining performance by enabling deep learning model to learn as humans do
 
■ LG AI Research, established after inauguration by Koo Kwang-mo, the chairman and CEO of LG Corp. strengthens AI competitiveness by jointly investing 20 billion won in SoftBank Ventures’ AI Investment Fund
 
 
LG AI Research, LG's AI-dedicated organization, released its first research achievements since its establishment at a conference held by Association for the Advancement of Artificial Intelligence (AAAI), the most prestigious global AI society.
 
The LG AI Research presented two research papers at the conference on the topics of “Explainable AI” and “Continual Learning,” which were jointly conducted with the University of Toronto in Canada.
 
Each year, numerous AI research institutes from around the world attend AAAI to publish research papers, where an opportunity is provided to estimate the AI competitiveness of participating countries. Hence, being selected by AAAI is a significant achievement as the research content and its technological value are globally acknowledged.
 
In 2018, LG established the Toronto AI Research institute with the University of Toronto, which holds world-class AI research achievements and technologies.
 
The two papers published this time reported the results of research conducted in collaboration with the University of Toronto.
 
Explainable AI is a technology that not only simply informs the results, but also logically explains how AI derived the results in a way that humans can easily understand.
 
For instance, when analyzing an X-ray image with AI, Explainable AI not only addresses the presence of an abnormality of a specific body part, but also explains its judgment in detail based on the given image.
 
In collaboration with a research team led by Professor Konstantinos Plataniotis at the University of Toronto, the has focused on the research of Explainable AI technology to improve the accuracy and fidelity of its explanation compared to that of existing technologies, so that it can be adopted in image recognition.
 
Explainable AI is a key technology in AI development that aids or replaces human decision-making in fields where reliability is vital, such as healthcare, finance, and law.
 
Furthermore, as research conducted even by a global company such as Google is still in a preliminary stage, AI development is a field that has the potential for South Korea to become the ‘First Mover’.
 
Continual Learning refers to the sequential learning of data by AI. For example, it allows a comprehensive judgment based on the newly learned contents by storing important data from the past, utilizing its short-term memory as a human does.
 
Previously, in terms of AI learning, the most significant challenges were the rapid increase in memory usage when working with a large amount of data and the reduction in accuracy when using an inadequate amount of data.
 
The publication by the LG AI Research presents a technology that maintains or improves performance while reducing the memory usage by AI when learning.
 
The LG AI Research and the team at the University of Toronto, led by Professor Scott Sanner, developed a technology that improves learning performance by up to 40% compared to the conventional methods; this was achieved for the first time by applying the Shapely value, which measures the importance of data to continuous learning.
 
The joint research team also won the “Continuous Learning Technology Contest” hosted for the first time by the globally renowned international conference, Computer Vision and Pattern Recognition (CVPR), among 79 other teams, including Amazon, China Academy of Sciences, and Tokyo University.
 
“Our goal is to advance AI technology research so that customers can directly experience the development,” said Kyunghoon Bae, head of LG AI Research. “We will continue to recruit young AI talents and cultivate 1,000 AI experts by 2023 to secure the world's best AI competitiveness,” he added.
 
Meanwhile, LG AI Research was established in December 2020 as LG's think tank for securing the latest global AI source technology and addressing various AI-related challenges. It recruited global AI scholar Honglak Lee, a professor at the University of Michigan, as a Chief Scientist of AI(CSAI), and it is continuously working on creating a global AI ecosystem through cooperation with Seoul National University, Toronto University in Canada, and other AI research institutes from around the world.
 
In addition, four subsidiaries of LG, including LG Electronics, LG Chem, LG Uplus, and LG CNS, have been rigorously investing jointly in AI with approximately 20 billion won in a fund created by SoftBank Ventures, totaling to an amount of approximately 320 billion.