Development of Ultra-Large AI That Understands User Emotions Like Humans
KT Plans to Evolve All AI Services to the Next Level
Bae Sun-min, head of the AI2XL Research Center at KT Convergence Technology Institute, is presenting the research direction of 'KT AI 2.0'.
[Asia Economy Reporter Lim Hye-seon] KT will commercialize AI services based on ultra-large-scale artificial intelligence (AI) within this year. Ultra-large-scale AI refers to AI designed to think autonomously like humans by learning from massive amounts of data, supported by infrastructure capable of large-scale computations.
On the 19th, KT held the 'KT Digico Study' at the KT Convergence Technology Institute in Seoul and announced on the 22nd its research direction called 'AI 2.0,' which aims to evolve all KT AI services to the next level based on ultra-large-scale AI.
During the 'KT Digico Study' event, Bae Soon-min, head of the AI2XL Research Center at KT Convergence Technology Institute, conducted a town hall interview under the theme 'Interaction AI Core Technology,' introducing KT's development of voice intelligence, language intelligence, and visual intelligence based on ultra-large-scale AI. He also revealed plans to apply these technologies to service areas such as AI Contact Center (AICC), Giga Genie, Genieverse, AI robots, and next-generation intelligent transportation systems (C-ITS).
The Core of Ultra-Large-Scale AI is Empathy
Director Bae emphasized that the differentiating direction of AI KT aims to develop is 'empathetic AI.' He explained, "The evolved KT AI will learn the surrounding and contextual situations autonomously and intervene appropriately at moments when users need it."
To realize more advanced AI, KT is developing ultra-large-scale AI models through multilateral joint research within the AI One Team. KT and the AI One Team plan to expand infrastructure scale to support models with over 200 billion parameters in the future.
The research team prioritizes eliminating AI bias by strengthening training data filtering and balancing data, while actively utilizing various deep learning-based detection technologies such as 'toxic content' filtering to sufficiently address AI ethics issues in advance.
In particular, KT's ultra-large-scale language model is designed in a 'task-oriented' form ready for immediate commercialization. Enhanced elemental technologies such as 'speech-to-text automatic conversion' and 'text analysis' through the ultra-large-scale language model will first be used to advance KT's AI services currently in commercial use, such as Giga Genie and AI Contact Center (AICC), and will rapidly apply to various business sectors of AI One Team participating companies in manufacturing, finance, logistics, and distribution.
Especially, once the 'Korean language model using ultra-large-scale AI,' which domestic and international major companies like Google are researching, is commercialized, it can be utilized in various fields.
Director Bae explained that the Korean language model evolved based on KT's ultra-large-scale AI will be the first AI equipped with 'empathy,' the most significant feature in human interaction.
According to Director Bae, KT AI will understand human language and generate language based on data through deep learning. It can infer context to comprehend incomplete sentences even in noisy environments, recall past conversations to communicate with users in simple language while conveying emotions, and analyze situations or actions using videos or photos to interact with humans. He also explained KT's plan to implement human-centered AI through convergence with humanities and cognitive science and empathetic UX based on joint research with KAIST.
KT Advances All Artificial Intelligence Services to the Next Level
Furthermore, KT unveiled its 'AI 2.0' vision to commercialize ultra-large-scale AI within this year and evolve all KT AI services to the next level through it.
In the business-to-consumer (B2C) sector, KT plans to develop next-generation AICC capable of empathizing with users' emotions. It will also develop 'multi-turn expert consultation' services enabling AI to conduct continuous conversations like humans in specialized fields such as childcare and law. Additionally, KT plans to innovate the conversation quality of Giga Genie, currently used by 3.3 million users.
In the business-to-business (B2B) market, KT aims to improve the conversation quality of KT AICC, which is spreading mainly in the financial sector, and implement services optimized for various industries. By minimizing simple response labor of workers in industrial sites, it is expected that AI-applied fields can increase industrial productivity by more than 30%. KT also announced a business goal to achieve over 180 billion KRW in AICC orders this year by utilizing such advanced AICC solutions.
Moreover, based on ultra-large-scale AI and KT Cloud, KT will create an open research and development (R&D) ecosystem with the AI One Team and open an AI R&D portal called 'GenieLabs.ai' to expand KT AI into various industries such as finance, vehicles, robots, and logistics. KT's ultra-large-scale AI will grow together with other companies and institutions through the cloud.
Bae Soon-min, head of the AI2XL Research Center, stated, "Research and development of AI technology require comprehensive integration with actual business sites and users above all else," adding, "KT is a unique company that generates actual revenue and achieves a virtuous cycle of AI technology and services." He continued, "Together with KT's creative researchers, we will develop empathetic AI technology that can embrace various generations."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
![Clutching a Stolen Dior Bag, Saying "I Hate Being Poor but Real"... The Grotesque Con of a "Human Knockoff" [Slate]](https://cwcontent.asiae.co.kr/asiaresize/183/2026021902243444107_1771435474.jpg)
