Digital Government Innovation, Preparing for Hyperscale AI and Cloud

DEC 16, 2024 Eunju Hong

Since the concept of artificial intelligence was only established in 1956, it astonished the world when AlphaGo won its big game in 2016. In 2022, the introduction of ChatGPT sparked the discussion around "generative AI," which has now become part of our daily lives to a remarkable extent. Recently, a demonstration video was released showcasing the robot "Figure 01," created through a collaboration between the humanoid robot development startup "Figure AI" and the developer of ChatGPT, "OpenAI." In the video, when a person says, "Give me something to eat," the robot understands that the only edible item on the table is an apple and hands it to the person. This is how the future, where AI autonomously makes decisions and takes actions based on human queries and instructions, is rapidly approaching.

As we face the emergence of new technologies such as "generative AI" and "hyperscale AI," this article examines how the adoption and transition to cloud technology for digital government innovation have progressed and what needs to be prepared in 2024.

The Journey of Digital Government Innovation

In this digital transformation era, the government has been promoting a "Cloud-First Policy," encouraging the adoption of cloud technology in public institutions. This journey, led by the Ministry of the Interior and Safety of South Korea (MOIS), has seen various phases, from institution-specific initiatives to adopting private clouds amid changing environments.

(1) Cloud Migration/Integration Projects

In 2021-2022, many public institutions pushed forward with cloud infrastructure migration projects funded by the MOIS. By 2023, under government policy, not only were these institutions adopting cloud infrastructure, but each ministry also developed a roadmap for transitioning to cloud native systems. 2023 marked the beginning of serious preparations to convert approximately 10,000 government systems to cloud native by 2030. Thus, 2024 is considered the inaugural year for the widespread adoption of hyperscale AI and expanding cloud native systems across the government.

Progress in Cloud Migration by Public Institutions

The MOIS has developed various 'adoption models' to accelerate and appropriately implement cloud technology within public institutions. One of the most notable examples is the "Private Cloud Center (Private Model)." The private model involves using the 'Public Cloud' infrastructure (service) provided by private cloud service providers certified under the Cloud Security Assurance Program (CSAP), such as Samsung SDS, to transition, build, operate, and manage the business systems of public institutions. This setup is logically divided into a "DMZ Zone" for internet access and a "Private Zone" for internal business operations. While this model enables rapid transitions to cloud services, it can only be used for general business operations with lower security requirements, excluding systems that handle sensitive information related to national security. As of 2023, a total of 5,465 systems, such as those used for introducing institutions or providing services to the public, have transitioned based on this model. One important point to remember is that public data security is a critical issue. In February 2024, the Ministry of Science and ICT of South Korea (MSIT) announced administrative guidelines for "medium and high security level assessments" for business systems. Therefore, public institutions must assess and prepare the appropriate security measures for each system based on its security level.

The second model is the "Public Cloud Center (Public-Private Partnerships)." This model involves collaboration between private data centers and administrative or public institutions to build a "Private Cloud" tailored to the institution's specific needs. It uses private/dedicated cloud services provided by private cloud providers, and it allows relevant agencies and startups involved in local administration to use the cloud environment, making it a model for advancing digital innovation in many local governments. A prominent example is the pilot project conducted by Jeonbuk Provincial Government in 2022, which collaborated with Samsung SDS. This model involved using the 'Samsung Cloud Platform' to create a physically dedicated cloud and transitioning sensitive internal systems to the cloud.

Public Cloud – Public-Private Partnerships

공공 클라우드 센터 - 민.관 협력형

민간 주도형, 민간 구축형을 설명하는 테이블
	민간 주도형(설비협력)	민간 구축형(설비+설비협력)
민감정보 시스템	지차체/공공기관 데이터센터 On-Site Cloud (전용), Private 지차체/공공기관 건물운영 보안관제 AP운영	민간 데이터 Private Cloud (전용), Private 지차체/공공기관 보안관제 AP운영
민감정보 시스템	On-Site Cloud (전용), DMZ 민간 클라우드 서비스 제공자 설비운영	Public Cloud (공유) 민간 클라우드 서비스 제공자 건물운영 설비운영

특장점

행정부 클라우드 컴퓨팅 서비스 활용 모델 시범사업(~22년)

민간 주도형 모델(2): 세종시
민간 구축형 모델(3): 전남도청,광주시청,경남도청,제주시청

Private Dedicated Cloud 서비스

(2) Digital Service Specialized Contract System

Building on these cloud adoption models by the MOIS, public institutions have rapidly advanced their cloud migration since 2021 through the "Digital Service Specialized Contract System." This system allows contracts for desired digital services to be processed online through a specialized platform. As of March 5, 2024, a total of 1,182 contracts worth 448.3 billion KRW have been executed through this system. Previously, it took over 80 days to plan a project, issue a request for proposal, and finalize service contracts. However, with cloud computing service providers like Samsung SDS registering their services through the MSIT's "Support System," public institutions can now select the desired services from the "Digital Service Mall" managed by the Public Procurement Service and complete the contract process in about two weeks. In 2021, the Digital Service Specialized Contract System was used in numerous projects, including large-scale contracts like the High-Performance Computing Resource Lease Service by the National IT Industry Promotion Agency of South Korea. This has enabled the rapid adoption of various digital and innovative technology services.

Digital Service Specialized Contract System

Digital Platform Government

Until 2023, the government focused on building a cloud infrastructure environment. As mentioned earlier, 2024 will mark the beginning of the widespread adoption of hyperscale AI (generative AI) and the expansion of cloud native systems across the government.

Have you heard the term "Digital Platform Government?" It may sound familiar. The Digital Platform Government aims to create an environment where systems and data from various agencies can be interconnected and shared through a digital platform. This initiative is pushing for private sector collaboration to swiftly integrate private technology into government-led projects. The concept also involves transforming the infrastructure environment to use AI and data. In other words, digital government innovation will be driven by the "continuous pursuit of cloud native solutions, supported by innovative technologies like digital platforms and hyperscale AI." Achieving this will require strong government backing and substantial budget allocations. This year's budget for the Digital Platform Government (DPG) has been finalized at 938.6 billion KRW, a 123% increase from the 420.7 billion KRW allocated in 2023. The budget will be distributed across four key areas: "Government for the People," "Smart One-Team Government," "Private-Public Growth Platform," and "Trustworthy and Secure DPG Implementation." Each area will have specific tasks defined and executed accordingly.

Directions of the Digital Platform Government

Preparing for Hyperscale AI

The core of the Digital Platform Government's initiatives is "Hyperscale AI." The point is on how AI will be adopted and integrated. In April 2023, the "Digital Platform Government Implementation Plan" was announced, which includes initiatives for building the necessary technology and industrial infrastructure, fostering an innovation ecosystem, and establishing supportive systems and cultural norms for hyperscale AI. Each institution focuses on government-driven initiatives and is gradually preparing by undertaking PoC or consulting-based concept validation projects to develop public and administrative hyperscale AI services.

Hyperscale AI Use Cases

Use Case 발굴 예산준비 과제수행 개별사업 PoC(proof of concept)

기관명	개념검증 과제
고용노동부	노동위원회 데이터 기반의 조사보고서 작성지원 시스템 실증
관세청	해외직구 등 수출입컨설팅(FTA 협정 포함) 등 AI 직구비서 서비스
국립중앙도서관	도서관 데이터를 학습한 도서관 서비스 질의응답 챗봇 서비스
국민권익위원회	국민 고충 민원의 민원 의도를 파악, 자동 답변추천 및 자동처리 실증
국민연금공단	국민연금 관련 다양한 민원에 대한 AI 기반의 대민응대 챗봇 서비스 실증
국세청	세법, 상담 등 업무 수행을 지원하는 비대면 업무 도움 서비스 실증
국회도서관	국회의 다양한 의회 정보와 뉴스 및 소셜미디어 분석, 알림 서비스 실증
근로복지공단	산재보상 실시간 상담서비스 지원 및 산재 신청 서비스 테스트 및 실증
법무부	범죄예방업무를 위한 문서제작에 사용되는 입력 값 검증하는 서비스 실증
서울교통공사	도시철도 안전관련 데이터 학습한 AI가 현장근로자 대상 해결방안 등 제공
소상공인시장진흥공단	계약 등 규정 가이드를 기반으로 행정업무를 지원하는 서비스 실증
수원시청	지자체 업무 관련 검색, 요약, 분석, 챗봇 등 행정지원 서비스 실증
양산시청	공문서 기반의 사업계획, 보도자료 등 초안 작성 및 요약 정리 서비스 실증
인천교통공사	인천 도시철도 시내버스, 장애인 콜택시 등 실시간 민원 상담 서비스
조달청	사업계획 및 제안요청서 작성을 위한 RFP 초안을 생성하는 서비스
중소기업중앙회	정책 건의, 사후관리 등 업무 효율화 목적의 민원 응대 챗봇 서비스 실증
한국건강증진개발원	기 구축된 흡연예방 및 금연실천 콘텐츠 기반의 QA 서비스 테스트
한국공항공사	공항이용 정보를 예측 추천하고, 관련 편의시설등을 안내하는 서비스
한국관광공사	기관 보유 서비스에 AI를 적용, 사용자 맞춤형 관광지, 여행지 검색 서비스

(1) Use Cases and the Government’s Preparedness

What are some AI services tailored for public sector tasks? Representative applications include administrative task management, welfare issue response, and handling public inquiries. For example, in the National Assembly Library, which consists of four affiliated institutions, records produced by these institutions are transferred to the library. By applying AI to these transferred records, an intelligent search system can be developed, making it easier to access related data. In the case of the Ministry of Government Legislation, AI can be applied to court or investigation records to obtain summarized information. Using guidelines or training data, providing services like responding to 119 emergencies or citizen inquiries through interactive chatbot services is possible.

The government has already begun laying the groundwork for hyperscale AI by developing pilot services through public-private partnerships, such as the "AI-powered official document drafting" service. In January 2024, the government announced the "Roadmap for Hyperscale AI," kicking off numerous projects. One of the most notable is the "Development of a National Large Language Model (LLM)," which aims to implement a government-wide LLM by 2025. This model will enhance service satisfaction and drive innovation in civil servant tasks. The government's vision goes beyond merely incorporating AI into individual tasks, such as civil complaint services. Instead, the goal is to create an environment where hyperscale AI, based on shared data across institutions, unifies public portals, provides personalized services, and fosters collaboration and innovation across ministries.

(2) Hyperscale AI Adoption Strategy

What needs to be prepared for introducing hyperscale AI? Since 2023, Samsung SDS has identified over 200 use cases through the needs of Samsung affiliates and external clients. These use cases include both common applications and those specific to particular tasks. To effectively introduce hyperscale AI, it is crucial to first identify which tasks require its application. The focus should be on selecting use cases for tasks that should be prioritized and where efficiency can be maximized. During this process, the essential data needed for implementation will be identified. The key to successfully implementing and applying hyperscale AI lies in the "data." It is vital to determine what data the hyperscale AI will learn from and how it will be used in the chosen use cases. Therefore, it is essential to organize and classify existing data by type and transform it into a usable format while considering its shareability with other organizations.

But how do you build a generative AI service? Most public sector operations are within the administrative network, raising concerns about using data within this network or linking it with external data. In June 2023, the National Intelligence Service issued security guidelines for using generative AI. To build a service using generative AI based on APIs like ChatGPT, it must follow guidelines ensuring data security and system security, including data anonymization and API key management to guarantee data validity and confidentiality. However, this is a very challenging task. Thus, it is necessary to use an AI platform that supports high-efficiency/high-security infrastructure, integration of internal data/systems with external systems, various LLM integrations, optimized serving, user portals or chat services, and the development of Copilots for legacy systems. Samsung SDS offers the generative AI service platform FabriX to facilitate the adoption of hyperscale AI.

Samsung SDS FabriX, the Generative AI Service Platform

Hyperscale AI services can be configured using an AI platform in various architectures. Private Configuration: In this setup, a private LLM and AI platform are placed within the internal network, allowing all data to be processed internally. This requires GPU infrastructure for fine-tuning and serving the private LLM, which minimizes the risk of external data leakage. Public Configuration: This setup uses external public LLMs, such as OpenAI's GPT or HyperCLOVA X, along with an AI platform to leverage the organization's data. This approach involves costs for using the LLM and other services, and it requires defining how internal data will be transferred to external AI platforms. However, most public institutions require a hybrid setup. In other words, the AI platform is commonly located within the internal network, allowing for the choice between a private LLM and an externally linked public LLM through APIs. This configuration can use internal data via Retrieval-Augmented Generation (RAG) technology.

Generative AI Architecture

Public and administrative institutions adopting hyperscale AI must consider various aspects, such as identifying use cases, designing cloud infrastructure and AI platform-based architecture, and defining and preprocessing data. Therefore, it is advisable to work comprehensively with a specialized partner who can provide consulting services, conduct technical verification and design, build the service, and support the service's operation and expansion throughout its entire life cycle.

Preparing for Cloud Native

By 2023, administrative and public institutions were required to apply cloud native technology to existing or new systems, except in cases where it was unavoidable.

(1) Government Project Implementation Plan

Cloud native is considered the highest stage of cloud maturity, involving the development/execution of applications using cloud-based containers and microservices architecture (MSA). In May 2023, the MOIS instructed each institution to independently establish and implement a cloud native migration roadmap for their systems. The goal is to achieve 50% cloud native conversion for existing systems and 70% for new systems by 2026. Also, in January 2024, the government organized an informational session to accelerate the cloud native migration. During this session, business plans submitted by various institutions were reviewed and selected, followed by the announcement of budget support for the migration to cloud native systems.

Government-Wide Common Cloud Platform

(2) Cloud Native Adoption Strategy

In this market environment, administrative and public institutions must prepare and assess their resource portfolios, including hardware, software, and applications, to migration to cloud native systems. The cloud migration follows the phases of "Planning for Use - Preparation for Adoption - Adoption/Use," and institutions should refer to the government's cloud migration guidelines throughout the process. In the "Planning for Use" phase, institutions should collaborate with cloud partners like Samsung SDS to discuss which systems to transition and by when, including infrastructure considerations. A key factor in this phase is the obsolescence of resources. Institutions should focus on rapidly transitioning outdated resources and prioritizing tasks that require improvement, all while developing an integrated migration plan that includes operational considerations. In the "Preparation for Adoption" phase, security, stability, scalability, preliminary coordination with related agencies on information projects, cost estimation, and securing the necessary budget should be considered. In particular, when deciding on the implementation approach in the "Adoption/Use" phase, it is crucial to consider the government's direction towards digital innovation, especially under the Digital Platform Government initiative. Reviewing which government budgets are available for use to support various digital innovation projects could be beneficial during this phase.

Cloud Native Migration Process

Afterword

Adopting cloud native systems and hyperscale AI in the public sector is no longer optional but necessary. To select and implement the most suitable models for each public institution, a comprehensive approach is required. The digital transformation through cloud adoption and the application of hyperscale AI is now an integral part of institutional evaluations. Please work step by step toward successful digital transformation in collaboration with cloud and AI specialists like Samsung SDS, following the government's guidelines for hyperscale AI and cloud native migration.

▶ This content is protected by the Copyright Act and is owned by the author or creator.
▶ Secondary processing and commercial use of the content without the author/creator's permission is prohibited.

Eunju Hong

Strategic Marketing Team at Samsung SDS

Had been in charge of digital transformation in samsungsds.com, solution page planning/operation, based on her work experiences in IT trend analysis, process innovation, and consulting business strategy, and is now in charge of content planning and marketing through trend/solution analysis for each main business sector of SDS.

Use Case	일반 지식검색	사내 지식검색	결산리포트 작성	...
Gen AI Service	Portal	Chatbot	Copilot	...
System/Data		KMS	ERP	...
LLM Model	상용	오픈소스	업종 특화	...
Infra	Public	Private	On-prem.	...

Digital Government Innovation, Preparing for Hyperscale AI and Cloud

The Journey of Digital Government Innovation

공공 클라우드 센터 - 민.관 협력형

디지털 서비스 정의

계약방법 규종

전문시스템 운영

Digital Platform Government

Preparing for Hyperscale AI

구현 시 필요 요소

고려사항

GenAI Service Platform

Private

내부망

Hybrid

내부망

Public

내부망

Preparing for Cloud Native

클라우드 플랫폼

이용기획단계

도입준비단계

도입 및 이용단계

Afterword