ExecLeaders PDF Content

Page 1 Transcript GFX: LG DRIVING INNOVATION WITH AI AND HUMAN COLLABORATION HWAYOUNG (EDWARD) LEE: I believe that collaboration between AI and human experts will make a huge impact in our society. GFX: HWAYOUNG (EDWARD) LEE: VP, LEAD AI BUSINESS DEVELOPMENT UNIT LG AI RESEARCH HWAYOUNG (EDWARD) LEE: And the brand name of LG's super giant AI EXAONE, EXpert AI for everyONE, shows the way to use AI for innovations. Another example is to use multi-modal text image generation AI model for the product design. It is built on top of the AWS Sagemaker. From last year we have been developing EXAONE Atelier, a creative studio that could inspire human artists to designers, with LG Household and Healthcare consumer and the cosmetic part of the company inside LG. By providing EXAONE Atelier as a collaboration tool we successfully make a lot of good business use cases such as launching our tattoo printing devices in the global market. We found that today's tattoo is not just a permanent one. It could be a day to day fashion item. It's really hard to design well-made tattoos within a limited amount of time before launching. So by leveraging the EXAONE Atelier we provided some of the collaboration tools between the AI and the human. So they just put in some of the text messages to the AI model, and the AI we want to automatically generate some of the initial image, and then putting some of their human expertise on top of that generated image, 10,000 different tattoo ideas have been generated. AI can help the human expert in two different manners. The first one is to inspire them when they are doing some creative jobs. And the other one they would like to do some tedious stuff. AI is going to dramatically increase productivity. We would like to try both. Especially in the inspiration area we collaborate with Parsons Design School. Is it possible to inspire an expert human? Is it possible to inspire the expert human designers or artists and creators to be more creative based on the image generation model? So we successfully developed that in the marketing campaign of the cosmetic product. In LG Household and Healthcare that was a great success story. LG Household and Healthcare usually produce a huge amount of the cosmetic products. So whenever they try to do a new concept of their marketing campaign it is really hard to figure out some of the distinctive and differentiated marketing concepts. The marketing managers would like to utilize EXAONE Atelier to inspire their new marketing concept. So they just put in some of the unusual text. For example the hidden oceans. So some of the ingredients in our cosmetic product come from the ocean. So the AI will be automatically generated, this is the hidden ocean people would like to love. And then the marketing campaign, the experts and the human designers would like to find out wow this is a great image from the image generation model. And customers really really loved it because the marketing concept is so unique. Five years ago when there was a hit in the AI in Korean industry a lot of companies would like to adopt the AI technology in their businesses. One of the most easy areas that could leverage AI is business inspection. Most of the business leaders and LG affiliates would like to adopt the short term solution to maximise their current profits. However like other cases most of those approaches were not that effective because of the limitation of the supervised learning methodologies. So for the last three years LG AI Research has been collaborating with LG Innotek to develop the state of the art vision inspection AI technologies for the camera model which is the major part of the smart phones. So we identified the business issues from scratch and listed up the required AI technologies that could solve them fundamentally. The unsupervised learning continual learning or active learning and overt detection like ah state of the art AI technology has been developed under the masterplan of adopting those in the factory line. The easiest way to do the vision inspection is to build a customised AI model for each different factory and each different line and each different product. But when you spread that model to other factories or other products or other mass production lines it is very hard to do that. What is the real actual problem they have, how can we overcome that situation from scratch? So that is we think about another approach to unsupervised learning. Doesn't require any labelling. So we can dramatically decrease the cost. So the continual learning is very helpful to maintaining the accuracy even with the time passing by. Transcript LG driving innovation with AI and human collaboration S E R I E S N I N E

• Cover