The Evolution of AI Infrastructure and NVIDIA DGX Spark
Recently, the AI development environment has shown polarization due to the skyrocketing demand for high-performance GPUs. NVIDIA has released the DGX Spark, a compact supercomputer, to fill this gap. This device features 128GB of unified memory, enabling large language models to run on-device.
DGX Spark is optimized for simple training and inference tasks at the desktop level with low power consumption and compact size. It is especially suitable for companies and researchers aiming to reduce cloud dependency and operate AI models locally.
- Equipped with 128GB LPDDR5x unified memory
- Based on Grace Blackwell architecture with GB10 superchip
- Supports 5th generation tensor cores and FP4 data format
- Low power consumption up to 240W
- Connectable to laptops and local networks
- Ideal for running AI agents like OpenCLo
- Supports clustering of up to 4 units
- Stable operation of generative AI models
AI-Centered Security Automation and Zero Trust Strategy
At the 2026 World Security Expo, AI-powered security automation emerged as a key topic. AI-based autonomous SOCs automatically detect, analyze, and respond to security events, enhancing operational efficiency. Additionally, the zero trust security model has entered the full deployment phase, forming a new security paradigm beyond traditional uniform network separation policies.
AI technology is applied across various fields such as video security, access control, and network security, enabling real-time anomaly detection and automatic response. This reduces the workload of security experts and significantly strengthens security levels for enterprises and public institutions.
- Utilization of AI-specialized security agents
- Introduction of Autonomous Security Operation Centers (SOC)
- Implementation of Zero Trust Network Access (ZTNA)
- Expansion of integrated IT security platforms
- Advancement of AI-based Cyber Threat Intelligence (CTI)
- Enhanced physical security using acoustic AI and vision-language models
- Real-time security data collection and analysis
- Security strategies linked with National Network Security Framework (N2SF)
Smart Theme Park 9.81 Park: Innovating Fun with Data and AI
9.81 Park, developed by Daesung Fintech Monolith Division in Jeju, is a smart theme park themed around gravitational acceleration. It records and analyzes all user activities in real-time, offering challenges and rewards like a game. AI, IoT, RTLS, and video recognition technologies are utilized for attraction operation and safety management.
Especially, the introduction of a licensing system and competition framework fosters visitor return rates and fandom. AI edits driving data and videos to create personalized highlight clips, maximizing user experience.
- Implementation of spatial game park platform
- Operation of Race 981 master licensing system
- AI-based vehicle recognition and driving data analysis
- Real-time location tracking and safety management
- Video editing using generative AI
- Automatic adjustment of licensing criteria based on weather conditions
- Integrated data management for various indoor and outdoor activities
- Plans for domestic and international expansion underway
Customized 3D Printed Glasses: Prism’s Innovation
Prism is a startup that produces personalized glasses using 3D scanning and 3D printing technologies. It precisely analyzes facial contours and vision data to provide optimized frames and lenses. The smart factory “Prism Foundry” in Seoul’s Seongsu district produces up to 150,000 custom glasses annually with a small workforce.
Prism offers a differentiated user experience through eco-friendly production processes, AI-based automatic facial modeling, and collaboration with renowned lens companies for high-quality lenses. They also operate a store in Manhattan, New York, actively pursuing global market entry.
- Precise facial contour measurement based on 3D scanning
- AI automatic facial modeling and design optimization
- Advanced 3D printing reducing waste and carbon emissions
- Provision of customized progressive multifocal lenses
- Operation of integrated online and offline services
- Global store expansion including New York, USA
- Diversification of designs through collaboration with famous designers
- Preparation for new markets such as smart glasses
Digital Finance Systems and Exchange Rate Error Case
In March 2026, Toss Bank experienced a yen exchange rate error, resulting in transactions at about half the normal rate for 7 minutes. This was caused by an automated system calculation error and insufficient anomaly detection. Approximately 28 billion KRW worth of transactions were made, raising concerns about financial system reliability.
Toss Bank canceled and reclaimed erroneous transactions based on the Electronic Financial Transactions Act and terms of service, but consumer rights and legal issues have arisen during the process. This incident highlights the importance of balancing digital financial innovation with internal controls.
- Automation error in exchange rate announcement system
- About 28.4 billion KRW traded at abnormal rates for 7 minutes
- Transaction cancellation and recovery under Electronic Financial Transactions Act
- Potential consumer rights and legal disputes
- Expanded impact compared to past similar cases
- Need for internal control and anomaly detection processes
- Impact of real-time exchange rate sharing and automated processing
- Controversy over equal compensation for small and large traders
FAQ
- Q1: What is DGX Spark suitable for?
A1: It is suitable for performing inference and simple training of large AI models locally at the desktop level. - Q2: What is Zero Trust security?
A2: It is a security model that verifies all access without distinguishing between internal and external, based on the principle of never trusting and always verifying. - Q3: How is AI technology used in 9.81 Park?
A3: AI is applied to collect user driving data, video editing, safety management, and automatic adjustment of licensing criteria. - Q4: What is the process for making customized glasses at Prism?
A4: It involves 3D facial scanning, vision testing, AI modeling, 3D printing, and custom lens production. - Q5: What were the main issues in the Toss Bank exchange rate error incident?
A5: The main issues were abnormal exchange rate application due to automation errors, insufficient internal anomaly detection, and consumer rights problems during transaction cancellation.
References
- IT Donga. (n.d.). IT Donga. https://it.donga.com/
- (n.d.). RSS Feed. https://www.zdnet.co.kr/news/news.xml
참고문헌
- IT Donga. (n.d.). IT Donga. https://it.donga.com/
- (n.d.). RSS Feed. https://www.zdnet.co.kr/news/news.xml