Resolving AI Temporal Errors and Data Center Investment: Strategies for Ensuring AI Reliability and Expanding Infrastructure

~17 min read

Resolving AI Temporal Errors and Data Center Investment: Strategies for Ensuring AI Reliability and Expanding Infrastructure

The Problem of Temporal Errors in Large Language Models (LLMs) and the Importance of Ensuring Reliability

Recently, news emerged that a joint research team from KAIST and Microsoft Research developed a new evaluation technology capable of resolving ‘temporal errors’ in Large Language Models (LLMs). This marks a significant advancement in enhancing the reliability of information provided by Artificial Intelligence (AI). For instance, AI models like ChatGPT sometimes provide incorrect answers to questions such as, “Who was the minister inaugurated last month?” These issues arise when AI fails to accurately reflect the latest information or properly understand temporal contexts, significantly undermining AI trustworthiness.

Even in South Korea, LLM-based services are rapidly emerging, but these temporal errors hinder user experience and diminish service reliability. For example, a news search service might display outdated or irrelevant articles first when a user requests the latest news on a specific event. This forces users to spend more time finding the information they need, leading to lower satisfaction with the service. Furthermore, in the financial sector, while LLM-powered investment advisory services are available, providing recommendations based on past data that do not align with current market conditions can lead to investment losses. Therefore, strengthening AI models’ temporal reasoning capabilities and ensuring they accurately reflect the latest information is a critical challenge.

To address these issues, AI developers in South Korea are making various efforts. For instance, they are developing technologies to continuously update LLMs with the latest information through real-time data streaming or researching methods to enhance information accuracy by strengthening integration with external databases. Additionally, they are actively incorporating user feedback to correct temporal errors and continuously improve model performance. These efforts are expected to contribute to increasing the reliability of AI services and enhancing user satisfaction.

A Novel Approach Utilizing Temporal Databases

To solve these problems, the KAIST research team introduced ‘Temporal Database’ design theory, validated over the past 40 years, into AI evaluation. This technology leverages the temporal flow and relational structure of data to automatically generate various types of time-based problems using only the database, without requiring humans to manually create evaluation questions. This represents an innovative method that moves beyond traditional evaluation approaches, which merely check for correct answers, allowing for a more accurate assessment of AI’s temporal reasoning capabilities.

A temporal database is designed to systematically manage data change histories and accurately query data at specific points in time. By applying the advantages of temporal databases to AI evaluation, it becomes possible to assess how well an AI model understands temporal contexts and how accurately it can infer information about the past, present, and future. For example, if the stock price fluctuations of a particular company are stored in a temporal database, and an AI model is asked to predict stock prices for a specific period, the model can be evaluated as having excellent temporal reasoning capabilities if it accurately grasps the temporal flow and provides reasonable predictions considering relevant economic indicators.

This temporal database-based evaluation method can be usefully applied to objectively measure AI model performance and suggest areas for improvement. Furthermore, it is expected to enhance the reliability of AI models used in various fields, allowing users to utilize AI services with greater confidence. In South Korea, research and development into temporal database technology are actively underway, and the number of system implementations utilizing temporal databases across various industries is increasing. This trend is anticipated to positively impact the improvement of AI models’ temporal reasoning capabilities and the overall reliability of AI services.

The Importance of Enhancing AI Reliability and Its Application in Finance, Healthcare, and Law

For AI to be widely adopted across various sectors, earning user trust is essential. The accuracy and reliability of AI are particularly crucial in fields that support critical decisions, such as finance, healthcare, and law. Resolving temporal errors and strengthening AI’s temporal reasoning capabilities will enable AI to provide more trustworthy information and judgments, thereby playing a vital role in expanding the scope of AI applications.

In the financial sector, AI is utilized in diverse areas such as investment advisory, credit scoring, and fraud detection. However, if an AI model is trained solely on historical data or fails to accurately reflect the latest market conditions, it can lead to severe problems like investment losses, credit assessment errors, or failed fraud detection. Therefore, strengthening AI models’ temporal reasoning capabilities and ensuring they accurately reflect the latest information is paramount for securing the stability and reliability of financial services. For example, an AI-based investment advisory service must analyze real-time market data and comprehensively consider past stock price trends and economic indicators to construct an investment portfolio. If the AI model fails to properly understand the temporal context or incorporate the latest information, it could lead to erroneous investment recommendations.

In the medical field, AI is applied in various areas, including disease diagnosis, treatment recommendation, and patient monitoring. However, if an AI model is trained on historical medical records or fails to accurately reflect the latest medical technologies, it can lead to serious issues such as misdiagnosis, inappropriate treatment recommendations, or patient safety concerns. Therefore, strengthening AI models’ temporal reasoning capabilities and ensuring they accurately reflect the latest medical information is crucial for improving the quality of healthcare services and ensuring patient safety. For instance, an AI-based disease diagnosis system must comprehensively analyze a patient’s past medical history, family history, lifestyle habits, and the latest test results to make a diagnosis. If the AI model fails to properly understand the temporal context or incorporate the latest medical information, it could lead to misdiagnosis.

In the legal sector, AI is used in various domains such as legal advisory, case law analysis, and litigation prediction. However, if an AI model is trained on past precedents or fails to accurately reflect the latest legal amendments, it can lead to severe problems like incorrect legal advice, inaccurate case law analysis, or erroneous judgment predictions. Therefore, strengthening AI models’ temporal reasoning capabilities and ensuring they accurately reflect the latest legal information is paramount for securing the fairness and accuracy of legal services. For example, an AI-based legal advisory service must comprehensively analyze relevant legal provisions, precedents, and legal interpretations to provide legal advice. If the AI model fails to properly understand the temporal context or incorporate the latest legal amendments, it could lead to incorrect legal advice.

The Evolution of Generative AI and the Importance of Data Centers

The Rapid Proliferation of Generative AI and Its Societal Impact

According to the ‘AI Index 2026’ report published by Stanford University’s Institute for Human-Centered Artificial Intelligence (HAI), generative AI is transforming from a mere technological tool into a ‘General Purpose Infrastructure’ that permeates global economic structures, scientific research methodologies, labor markets, and national policies. Generative AI is spreading across society at a much faster pace than past internet or mobile revolutions, impacting various aspects such as productivity, value creation, power structures, and social inequality.

In South Korea, generative AI is also rapidly expanding and being utilized across diverse industrial sectors. For example, in content creation, AI is used to automatically generate various forms of content, including text, images, music, and video. In marketing, AI helps create personalized advertising content and automates customer service. Furthermore, in education, AI provides customized learning content tailored to students’ proficiency levels and manages their learning progress, while in healthcare, AI assists with disease diagnosis and develops personalized treatment plans for patients. As such, generative AI is bringing about transformative changes across society, fundamentally altering our lives and work methods.

However, the rapid proliferation of generative AI simultaneously raises various societal issues. For instance, the spread of AI-generated fake news or misinformation can cause social confusion, and AI-generated content can lead to copyright infringement problems. Additionally, concerns are growing that as AI automates more jobs, labor market instability will intensify, and social inequality may worsen, with segments of the population having limited access to AI technology being marginalized. Therefore, to maximize the positive aspects of generative AI and minimize its negative impacts, in-depth discussions and policy formulation on ethical and social issues are necessary alongside AI technology development.

Intensifying Competition in AI Data Center Construction and Elice Group’s Investment Attraction Case

With the rapid spread of generative AI, the demand for immense computing resources required for AI model training and operation has surged, intensifying competition in AI data center construction. Elice Group secured strategic investments from Dongkuk Holdings and GS Ventures, laying the groundwork for building a 1GW-class hyperscale AI data center. This represents a crucial step towards gaining a competitive edge in AI infrastructure and meeting the growing demand for AI computing.

AI data centers are critical infrastructure for AI model training, inference, data storage, and management. The importance of AI data centers that provide high-performance computing resources is further emphasized, especially as generative AI models require massive datasets and complex computations. Data center performance, including stable power supply, efficient cooling systems, and high network bandwidth, directly impacts the quality of AI services. Elice Group’s data center construction is expected to strengthen the competitiveness of South Korea’s AI industry and promote the development and proliferation of various AI services.

Beyond Elice Group, various other companies in South Korea are investing in AI data center construction. For example, Naver is building large-scale data centers for its own AI model development and service provision, while Kakao is expanding its data center investments for AI research and development and service expansion. Furthermore, telecommunication companies such as SK Telecom, KT, and LG Uplus are actively participating in AI data center construction, and cloud service providers are also targeting the AI service market through their data center initiatives. This intensifying competition in AI data center construction is a significant factor driving the growth and development of South Korea’s AI industry and is expected to become even fiercer in the future.

The Importance of Data Center Construction and Strengthening South Korea’s AI Industry Competitiveness

AI data centers are critical infrastructure for AI model training, inference, data storage, and management. The importance of AI data centers that provide high-performance computing resources is further emphasized, especially as generative AI models require massive datasets and complex computations. Data center performance, including stable power supply, efficient cooling systems, and high network bandwidth, directly impacts the quality of AI services. Elice Group’s data center construction is expected to strengthen the competitiveness of South Korea’s AI industry and promote the development and proliferation of various AI services.

Data centers are not merely spaces for storing and managing servers; they are core production infrastructures for the AI era. The performance of a data center directly influences the quality of AI services, including AI model training speed, inference accuracy, and service response time, making it a crucial factor determining the competitiveness of the AI industry. For example, AI-based autonomous vehicles require real-time perception and decision-making about their surroundings, necessitating fast response times from data centers. Similarly, AI-based medical diagnostic systems must analyze large volumes of medical data and provide accurate diagnostic results, making high-performance computing resources from data centers essential.

To strengthen the competitiveness of South Korea’s AI industry, it is essential to secure technological competitiveness not only in data center construction but also in various fields such as data center operation technology, AI model development technology, and AI service provision technology. Furthermore, developing technologies to reduce data center construction costs and enhance energy efficiency is also important. The government should strengthen the domestic AI industry’s competitiveness through active policy support, including assisting AI data center construction and funding research and development for AI technology. Additionally, multifaceted efforts are needed, such as expanding educational programs for nurturing AI talent and strengthening international cooperation for AI technology development.

Accelerating AI Adoption Across Industries

AI Transformation (AX) in Manufacturing and Hyundai WIA’s AI Education Case Study

Hyundai WIA is accelerating its AI Transformation (AX) by providing AI education to all employees. This exemplifies efforts within the manufacturing sector to actively adopt AI to enhance operational efficiency and create new value. Through AX Literacy training, employees understand the basic concepts and application principles of generative AI, enabling them to apply AI to their respective job functions. Furthermore, the company supports all executives and employees in directly utilizing generative AI in their work through ‘H-CHAT,’ an AI tool developed by Hyundai Motor Group. This signifies that AI is no longer solely the domain of specific departments or experts but is becoming a tool accessible to all employees.

The manufacturing industry can achieve various benefits through AI adoption, including increased productivity, improved quality, and cost reduction. For example, AI can automatically detect defective products on production lines and optimize manufacturing processes to enhance production efficiency. Additionally, AI can predict product performance through simulations during the product design phase and prevent design errors in advance, thereby shortening product development cycles. Furthermore, AI can analyze customer data to develop customized products and increase customer satisfaction. In this way, AI can bring about transformative changes across the entire value chain of manufacturing.

South Korean manufacturers are striving to enhance their global competitiveness through AI adoption. For instance, Samsung Electronics is building AI-based smart factories to boost productivity and improve quality, while LG Electronics is developing AI-powered smart home appliances to increase customer satisfaction. Furthermore, Hyundai Motor Company is developing AI-based autonomous vehicles and working to lead the future mobility market. The government is pursuing various policies to support AI adoption in manufacturing, including providing R&D funding for AI technology development. Additionally, multifaceted efforts are needed, such as expanding educational programs for nurturing AI talent and strengthening international cooperation for AI technology development.

Innovation in Autonomous Driving and Vehicle Experience: Bosch and Qualcomm Expand Collaboration

Bosch and Qualcomm are intensifying competition in AI-based automotive technology by expanding their collaboration into the ADAS (Advanced Driver-Assistance Systems) sector. The two companies are jointly developing next-generation ADAS solutions that enhance autonomous driving, connectivity, and personalization features, all based on the Snapdragon Digital Chassis platform. This strategic move aims to respond to the era of Software-Defined Vehicles and focuses on accelerating the transition to a centralized computing architecture that integrates various in-vehicle functions. Such collaboration will be a crucial turning point that reshapes the future of the automotive industry.

Autonomous driving technology can be described as the culmination of AI technologies. Autonomous vehicles perceive their surroundings, safely plan driving routes, and control the vehicle through AI-based perception, decision-making, and control technologies. AI acts as the eyes, ears, and brain of autonomous vehicles, making it a core element of autonomous driving technology. For example, AI-based object recognition technology analyzes data collected from sensors such as cameras, radar, and lidar to identify surrounding objects like vehicles, pedestrians, bicycles, and traffic lights. Furthermore, AI-based path planning technology comprehensively considers map data, traffic information, and surrounding object information to plan safe and efficient driving routes. Additionally, AI-based vehicle control technology manages vehicle movements such as acceleration, deceleration, and steering to ensure safe driving.

In South Korea, autonomous driving technology development is also actively underway, with major corporations like Hyundai Motor Company, Samsung Electronics, and LG Electronics investing heavily in it. Startups are also participating in autonomous driving technology development, leading innovation based on diverse ideas and technological capabilities. The government is pursuing various policies to support autonomous driving technology development, including providing R&D funding. Additionally, it is easing regulations for autonomous vehicle test operations and promoting infrastructure construction for autonomous vehicle commercialization. Through these efforts, South Korea is expected to strengthen its autonomous driving technology competitiveness and lead the future mobility market.

Innovating Election Campaigns with AI Music

In the age of Artificial Intelligence (AI), as the election campaign landscape rapidly reshapes around video, short-form content, and social media, there’s growing interest in music-based content that can convey both messages and atmosphere in a short amount of time. While traditional election-related music content often required significant time and cost for production, AI music can generate songs of various moods and formats relatively quickly. This makes it highly valuable in fast-paced digital campaign environments. Consequently, examples of AI music being used for content creation are emerging in the political sphere, with attempts continuing in various formats such as background music for rallies, promotional videos, short-form content, and online theme music.

AI music is a technology that utilizes AI to compose, arrange, and perform music. Based on learned musical data, AI can create new music or modify existing music according to user requirements. AI music offers advantages over traditional music production methods, such as saving time and cost, and enabling the easy creation of diverse musical styles. For instance, AI can learn the style of a specific genre to compose new music in that genre, or modify music to match the user’s desired mood or tempo. Additionally, AI can synthesize various instrument sounds to create new sounds or add new instruments to existing music.

AI music can be utilized not only in election campaigns but also in various fields such as advertising, film, and gaming. For example, AI can automatically generate background music suitable for an advertisement video or compose music that matches the atmosphere of a specific scene in a movie. Additionally, AI can create background music or sound effects for games to enhance immersion. AI music technology is rapidly advancing and is expected to be applied in even more diverse areas in the future. In South Korea, AI music technology development is actively underway, with startups leading innovation in this field. The government is pursuing various policies to support AI music technology development, including providing R&D funding.

Challenges and Opportunities in the AI Era

The Importance of AI Ethics and Social Infrastructure Development

The rapid advancement and proliferation of AI technology offer numerous opportunities but also present challenges such as ethical issues and a lack of social infrastructure.

AI technology has the potential to positively transform our lives and society, but it can also give rise to various ethical problems. For example, issues such as AI making discriminatory decisions, infringing on personal information, or reducing employment opportunities can occur. Therefore, alongside AI technology development, in-depth discussions and policy formulation regarding AI ethics are necessary. AI ethics can provide guidelines for AI technology development and utilization, helping to prevent and resolve ethical issues that may arise from AI. For instance, AI ethics can establish principles and standards needed to ensure fairness so that AI does not make discriminatory decisions, protect personal information securely, and build social safety nets for job displacement.

Developing the social infrastructure necessary for AI technology proliferation is also a critical task. AI technology requires various infrastructures, including data, computing resources, and human talent. However, South Korea’s current AI infrastructure is insufficient compared to developed nations, with data acquisition and AI talent development being particularly urgent challenges. The government must expand data openness and utilization for AI infrastructure development and broaden educational programs for nurturing AI talent. Additionally, multifaceted efforts are needed, such as providing R&D funding for AI technology development and strengthening international cooperation for AI technology development.

In the AI era, strengthening citizens’ capabilities to understand and utilize AI technology is also important. AI technology should no longer be the exclusive domain of experts but a technology that all citizens can comprehend and use. Therefore, efforts are needed to expand AI education and enhance citizens’ understanding of AI technology. Furthermore, creative ideas from citizens who can leverage AI technology to solve social problems and create new value should be supported. In the AI era, AI technology, ethics, social infrastructure, and citizen capabilities must develop harmoniously to ensure that AI technology has a positive impact on society as a whole.

• Finance:
  • Financial Services Commission
  • Bank of Korea
  • Korea Exchange
• IT: