KT Corp

Foundation Models

South Korea-based Upstage specializes in the development of LLMs. The company’s flagship, LLM Solar, is claimed to be fast, cost-efficient, and powered by the Depth-Up Scaling (DUS) technique with its ability to scale LLMs efficiently. The LLMs can be customized for various use cases, including healthcare, customer support, finance, insurance, ecommerce, law, and media. The model is available on-premise, and API integration through platforms like Amazon SageMaker JumpStart. The company also offers a Document AI solution, an AI-powered OCR (optical character recognition) technology to automate workflows and process unstructured data.In September 2024, the company released the preview of its large language model (LLM) <a href="https://sp-edge.com/updates/33519">Solar Pro</a> with 22 billion parameters.Key customers and partnershipsIn January 2024, Upstage partnered with SK Networks to expand globally and combine Upstage's private language model capabilities with SK Network’s en-core data management services.The company’s product, Document AI, caters to demand in the finance, healthcare, and legal industries, with notable clients like Samsung and Hanwha.In May 2024, <a href="https://sp-edge.com/updates/29352">Upstage signed a memorandum of understanding with Flitto,</a> a provider of translation platform, to create an AI-based dataset that can handle under-resourced Asian languages.In June 2024, Upstage announced a<a href="https://sp-edge.com/updates/30956"> multi-year Strategic Collaboration with Amazon Web Services</a> (AWS) to expand the global reach of its AI offerings (namely its flagship LLM, SOLAR) via AWS’s “ISV Accelerate” co-selling program.

Empowering Tomorrow's Innovations with Generative AI Infrastructure

Generative AI Infrastructure

Rebellions is a South Korean startup that develops AI system-on chips (SoC). The company's primary product offering is ATOM, an AI chip launched in 2023, designed to compete with industry leader NVIDIA. It claims to be the first neural processing unit (NPU) from South Korea to be utilized in a data center for LLM. Rebellions aims to provide domain-specific AI processors coupled with optimized software. The Atom chip is specifically engineered for computer vision and chatbot AI applications.&nbsp;The firm claims its AI chips are more energy-efficient, consuming ~20% of the power of an NVIDIA A100 chip when performing targeted tasks. This energy efficiency is achieved through the chip's focus on specific tasks rather than broad functionality. Rebellions' approach includes a decentralized programming model, which the company claims will enable customers to leverage powerful yet energy-efficient AI hardware with seamless software integration.As of <a href="https://sp-edge.com/updates/30407">June 2024</a>, Rebellions and Sapeon Korea, developer of the X330 AI chip, were set to merge to focus on NPUs.Key customers and partnershipsKT, one of South Korea's largest telecom operators, is an investor in Rebellions and a prospective customer for its products.&nbsp;In October 2023, Rebellions and Samsung Electronics partnered to develop a next-generation AI chip, “Rebel,” aiming to tap into the growing GenAI market.

Rebellions.ai creates AI accelerators that bridge the gap between core semiconductor architectures and deep learning algorithms. Its objective is to provide domain-specific AI processors together with optimized software.

Upstage AI is a provider of customized AI services designed to help businesses with artificial intelligence. The company delivers AI models based on the demands of the business to address the customer's core activities while also improving the efficiency and progress of the business process, allowing clients to solve problems and meet the company's AI innovation goals.

With the mission to provide the most effective education for all, QANDA(Mathpresso) operates the world’s leading learning platform, QANDA, an AI-based learning app that allows students of all levels to receive instant answers and customized learning sources. Recognized for its technological advances, QANDA(Mathpresso) was selected as GSV Edtech 150, national representative of Google AI for Social Good APAC event in 2018, and guest speaker at AWS Public Sector Summit ASEAN in 2019. Since 2015, QANDA(Mathpresso) has raised over $105 million in total from investors including SoftBank Ventures Asia, Legend Capital, Mirae Asset Investment, GGV, and Goodwater Capital. Headquartered in Seoul, QANDA(Mathpresso) has offices in Tokyo, Hanoi, Jakarta, and Bangkok.

Sonde Health is a digital medicine company developing voice-based technologies to transform the way we monitor and diagnose our mental and physical health.

Table Manager is a reservation management solution that systematically manages telephone reservations.

Digital Pharm is a pharmaceutical company that specializes in electroceuticals and DTx.

NeuroSigma, Inc. is a medical device company based in Westwood, California. NeuroSigma is focused on commercialization of its unique Trigeminal Nerve Stimulation (TNS) technology, and has a growing pipeline of treatments targeting epilepsy, depression, attention deficit hyperactivity disorder (ADHD), and post-traumatic stress disorder (PTSD). NeuroSigma is actively engaged in cutting-edge preclinical research and development, as well as groundbreaking Phase I, II, & III clinical trials.

Banksalad is a fintech company that develops financial applications.

One Store is a mobile app market that allows users to download apps that are limited in Korea only. It is run by mobile carriers SK Telecom, KT, LG Uplus and tech company Naver. It has combined its own four different platforms for Android-based smartphones and other mobile devices. Users can upgrade one of the apps to use One Store. It can be used on Android phones and tablets in Korea. It provides games, apps, mobile commerce, e-books, and video content to 35 million members.  It was founded in 2016 and is headquartered in Seongnam Gyeonggi-do, South Korea.

HYUNDAI ROBOTICS OF NORTH AMERICA, in partnership with HYUNDAI HEAVY INDUSTRIES offers a complete line of automation robots to today’s highly competitive manufacturing marketplace. HYUNDAI has designed and built robots and robotic systems for their automotive group for over fifteen (15) years. The robotic equipment is designed and built with accuracy and durability in mind. Complying with automotive quality and manufacturing standards, HYUNDAI produces some of the most reliable robots within the industry.

Wochu.cn is an eCommerce platform for fresh vegetables.

Ustream provides a live interactive broadcast platform.     Its platform enables to view and broadcast content, including high school sporting events and red carpet movie premieres, political events, concerts, conferences, talk shows, interactive games, and personal milestones. The company also offers advertising and digital campaign solutions; solutions for mobile devices, streaming players, connected TVs, and other devices; production and event services such as event production, event and program development, account management/tech support and performance metrics, transmission and connectivity, and custom mobile applications services; and broadcaster services.     The company was founded in 2007 by John Ham, Brad Hunstable, and Gyula Feher. Its clientele and partners include Cisco, Sony, Discovery Communications, Georgetown University, and Samsung.     Based in Francisco, California, Ustream has additional offices in Los Angeles, California; Tokyo, Japan; and Budapest, Hungary.

Expanding 4.5 million fixed lines to 20 million in just 12 years, kt introduced universal telephone service to every citizen of Korea, leading the development and advancement of communications services.   More recently, kt established an advanced broadband network for the first time in Asia, and launched Korea's first communication satellite 'Mugunghwa (Sharon's Rose)', contributing to making Korea one of the most advanced countries in information and communications.  Since becoming a government-funded corporation in 1997, kt changed its business portfolio to focus more on wireless and internet services instead of fixed lines.   Taking over Hansol M.com and merging with its subsidiary KTF, kt expanded its business scope and included mobile communication as its main service.   kt became a privately owned company in 2002 and began concentrating on building broadband network. As a result, the company set several new records including the increase of ADSL subscribers to 6 million in just 3 years and played a leading role in the age of digital networking.

<div> KT Corporation (NYSE: KT) is South Korea's largest integrated telecom and digital platform service provider. Founded in 1981 as Korea Telecommunications Authority, KT was privatized in 2002 and has since expanded its operations beyond traditional telecommunications services. KT offers a wide range of products and services including: <ul> <li> Mobile voice and data services based on 5G, 4G LTE and 3G technologies </li> <li> Fixed-line telephone and broadband internet services </li> <li> IPTV and satellite TV </li> <li> Cloud computing and data center services </li> <li> Artificial intelligence and Big Data analytics </li> <li> internet of Things (IoT) solutions </li> <li> Digital financial services </li> </ul> KT has been at the forefront of South Korea's telecommunications advancements, playing a key role in the country's 5G rollout. As of June 2023, over 40% of KT's mobile users were on 5G networks. The company has invested heavily in artificial intelligence, pledging to invest KRW 7 trillion (USD 5.3 billion) in AI research and development by 2027. KT aims to leverage its AI capabilities across its media and content businesses, including content production and personalized TV services. The company is also exploring international expansion opportunities, particularly in spreading Korean pop culture content globally through partnerships and content remakes. </div>

KT Corp Q3 Profit Jumps 44%, Plans Digital Transformation

Foundation models that can handle a range of natural language processing (NLP) use cases. Typical tasks include generating summaries, translating, answering questions, and classifying text. These models typically use text inputs and generate text outputs.

Fine-tuned versions of large language models, which are focused on specific tasks and applications (e.g. code generation, medical event prediction, financial NLP tasks, etc). These are usually developed by training LLMs using custom data and also serve as foundation models that can be developed even further. These models typically use text inputs and can generate text and/or code outputs.

Foundation models that translate visual data based on features and contextual information identified during training. This enables models to interpret images and video, applying those interpretations to predictive or decision-making tasks. Typical tasks include image recognition and classification into categories. These models typically use image and/or video inputs and generate text outputs.

Foundation models that are trained to understand and work with two or more forms of input data (e.g., text and images) to solve problems and complete tasks using both natural language processing and computer vision. These models can also generate outputs across textual and visual modalities. Typical tasks include visual question answering, image captioning, and generating images based on text descriptions. These models typically use text and/or visual inputs and generate text and/or visual outputs.

Foundation models that use big data to suggest or recommend additional products to consumers. These models are designed to predict user-item interaction probabilities and provide recommendations to users. The models are trained to make these inferences based on historical data on customer interactions and non-interactions. These models typically use user features, item features and user interactions as inputs, and generate numeric values like ratings predictions and matching probabilities as outputs.

Foundation models that are capable of automatic speech recognition, transcription in multiple languages, and translation from these languages to English. These are sometimes referred to as audio models. These models typically use speech/audio as inputs and generate text outputs.

Foundation models that transform user text prompts into video clips.

Foundation models that generate music from text descriptions.

INCUMBENTS_PRESENT

PROMINENT_CATEGORY

<div>
 Korean telecommunications provider, KT, and Microsoft have forged a partnership to create a customized version of ChatGPT and other AI solutions specific to Korea. They plan on investing 1.8 billion USD over a span of five years dedicate to creating an AI model customized for Korea by the second quarter of 2025. In addition, they will develop specific cloud services for the public and financial sectors, while also establishing joint venture dedicated to AI transformation. KT expects to generate an additional 3.4 billion USD in sales over the next five years as a result of this collaboration.
 The partnership will result in the co-development of a tailored version of GPT-4o and a small language model based on Microsoft's Phi-3.5. These will be specialized for specific industries in Korea. KT is focusing on training these AI models on Korean data pertaining to education, history, and culture to increase their localization effectiveness.
</div>

KT Corp

Featured Industries

About EDGE

Contact Us

Resources

Area of Focus

Technology Research Made Effortless