• Journal of the Society of Disaster Information
• /
• v.19 no.2
• /
• pp.432-438
• /
• 2023

Purpose: The purpose of this study is to enable users to conveniently report risks by automatically classifying risk categories in real time using AI for images reported in the life safety prevention service app. Method: Through a system consisting of a life safety prevention service platform, life safety prevention service app, AI model serving server and sftp server interconnected through the Internet, the reported life safety images are automatically classified in real time, and the AI model used at this time An AI learning algorithm for generation was also developed. Result: Images can be automatically classified by AI processing in real time, making it easier for reporters to report matters related to life safety.Conclusion: The AI image automatic classification system presented in this paper automatically classifies reported images in real time with a classification accuracy of over 90%, enabling reporters to easily report images related to life safety. It is necessary to develop faster and more accurate AI models and improve system processing capacity.

• Electronics and Telecommunications Trends
• /
• v.35 no.3
• /
• pp.66-75
• /
• 2020

As the increasing expectations of a practical AI (Artificial Intelligence) service makes AI algorithms more complicated, an efficient processor to process AI algorithms is required. To meet this requirement, processors optimized for parallel processing, such as GPUs (Graphics Processing Units), have been widely employed. However, the GPU has a generalized structure for various applications, so it is not optimized for the AI algorithm. Therefore, research on the development of AI processors optimized for AI algorithm processing has been actively conducted. This paper briefly introduces an AI processor especially for inference acceleration, developed by the Electronics and Telecommunications Research Institute, South Korea., and other global vendors for mobile and server platforms. However, the GPU has a generalized structure for various applications, so it is not optimized for the AI algorithm. Therefore, research on the development of AI processors optimized for AI algorithm processing has been actively conducted.

• Electronics and Telecommunications Trends
• /
• v.35 no.7
• /
• pp.3-12
• /
• 2020

In this paper, we introduce ETRI AI Strategy #1, "Proactively Securing AI Core Technologies." The first goal of this strategy is to innovate artificial intelligence (AI) service technology to overcome the current limitations of AI technologies. Even though we saw a big jump in AI technology development recently due to the rise of deep learning (DL), DL still has technical limitations and problems. This paper introduces the four major parts of the advanced AI technologies that ETRI will secure to overcome the problems of DL and harmonize AI with the human world: post DL technology, human-AI collaboration technology, intelligence for autonomous things, and big data platform technology.

• Electronics and Telecommunications Trends
• /
• v.35 no.7
• /
• pp.13-22
• /
• 2020

There is no denying that computing power has been a crucial driving force behind the development of artificial intelligence today. In addition, artificial intelligence (AI) semiconductors and computing systems are perceived to have promising industrial value in the market along with rapid technological advances. Therefore, success in this field is also meaningful to the nation's growth and competitiveness. In this context, ETRI's AI strategy proposes implementation directions and tasks with the aim of strengthening the technological competitiveness of AI semiconductors and computing systems. The paper contains a brief background of ETRI's AI Strategy #2, research and development trends, and key tasks in four major areas: 1) AI processors, 2) AI computing systems, 3) neuromorphic computing, and 4) quantum computing.

• Journal of the Korean Society of Systems Engineering
• /
• v.19 no.1
• /
• pp.91-106
• /
• 2023

Multidimensional efforts such as budgeting, organizing, and institutionalizing are being carried out for the adoption of defense AI. However, there is little interest in eliminating duplication of defense resources that may occur during the AI adoption. In this study, we propose a theoretical conceptual model to optimize duplication of AI technology that may occur during the AI adoption in the vast defense field. For a systematic approach, the JCA of the US DoD and system abstraction method are applied, and the IMO logical structure is used to decompose AI requirements and identify duplication. As a result of analyzing the effectiveness of our conceptual model through six example defense AI requirements, it was found that the amount of requirements of data and AI technologies could be reduced by up to 41.7% and 70%, respectively, and estimated costs could be reduced by up to 35.5%.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.432-433
• /
• 2023

오늘날 AI 기술은 다양한 분야에서 활용되며 발전해나가고 있다. 하지만 AI 모델의 복잡도가 증가하며 AI의 산출 결과의 해석이 불가능한 Black-box 성격을 지니게 되었고, 이는 실 환경에서 AI 도입의 커다란 걸림돌로 작용하고 있다. 이에 따라 AI 판단 결과에 대한 Interpretation을 제공하는AI Decision Support의 중요성이 커지는 추세이다. 본 논문에서는 Reference 기반 Rule을 통해 AI 모델의 판단 결과에 대한 해석을 제공하고 입력된 데이터에 관한 Rule 적합도를 산출하여 AI Decision Support를 제공하고자 한다. 또한, Rule 적합도 정보를 기반으로 기존의 모델보다 정확한산출 결과를 통해 수집된 데이터의 Label을 확정시킨다. 이를 토대로 AI 모델의 업데이트를 실행하여 지속적으로 AI의 성능을 개선하면서도 지속 운용이 가능한 AI 운용 프레임워크를 제안한다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.75-79
• /
• 2023

문서에서 테이블은 중요한 정보들을 축약하여 모아 놓은 정보 집합체라고 할 수 있다. 이러한 테이블을 대상으로 질의응답하는 테이블 질의응답 기술이 연구되고 있으며, 이 중 언어모델을 이용한 연구가 좋은 결과를 보이고 있다. 본 연구에서는 최근 주목받고 있는 생성형 언어모델 기술을 테이블 질의응답에 적용하여 언어모델과 프롬프트의 변경에 따른 결과를 살펴보고, 단답형 정답과 생성형 결과의 특성에 적합한 평가방법으로 측정해 보았다. 자체 개발한 EXAONE 1.7B 모델의 경우 KorWiki 데이터셋에 대해 적용하여 EM 92.49, F1 94.81의 결과를 얻었으며, 이를 통해 작은 크기의 모델을 파인튜닝하여 GPT-4와 같은 초거대 모델보다 좋은 성능을 보일 수 있음을 확인하였다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.415-418
• /
• 2023

테이블 데이터는 일반적으로 헤더와 데이터로 구성되며, 헤더는 데이터의 구조와 내용을 이해하는데 중요한 역할을 한다. 하지만 웹 스크래핑 등을 통해 얻은 데이터와 같이 다양한 상황에서 헤더 정보가 누락될 수 있다. 수동으로 헤더를 생성하는 것은 시간이 많이 걸리고 비효율적이기 때문에, 본 논문에서는 자동으로 헤더를 생성하는 태스크를 정의하고 이를 해결하기 위한 모델을 제안한다. 이 모델은 BART를 기반으로 각 열을 구성하는 텍스트와 열 간의 관계를 분석하여 헤더 텍스트를 생성한다. 이 과정을 통해 테이블 데이터의 구성요소 간의 관계에 대해 이해하고, 테이블 데이터의 헤더를 생성하여 다양한 애플리케이션에서의 활용할 수 있다. 실험을 통해 그 성능을 평가한 결과, 테이블 구조 정보를 종합적으로 활용하는 것이 더 높은 성능을 보임을 확인하였다.

• International journal of advanced smart convergence
• /
• v.12 no.4
• /
• pp.154-163
• /
• 2023

With the rapid development of AI technology, ChatGPT and other AI content creation tools are becoming common, and users are becoming curious and adopting them. These tools, unlike search engines, generate results based on user prompts, which puts them at risk of inaccuracy or plagiarism. This allows unethical users to create inappropriate content and poses greater educational and corporate data security concerns. AI content detection is needed and AI-generated text needs to be identified to address misinformation and trust issues. Along with the positive use of AI tools, monitoring and regulation of their ethical use is essential. When detecting content created by AI with an AI content detection tool, it can be used efficiently by using the appropriate tool depending on the usage environment and purpose. In this paper, we collect data on AI content detection tools and compare and analyze the functions and characteristics of AI content detection tools to help meet these needs.

• Journal of Internet of Things and Convergence
• /
• v.8 no.5
• /
• pp.71-76
• /
• 2022

The innovative development of the 4th industry and the COVID-19 pandemic caused a great change in the education, eventually requiring elementary, middle and high schools, including kindergartens, to implement artificial intelligence(AI) education. However, since early childhood AI education is conducted in the form of results-oriented and special activities, the need for research on what early childhood AI education is and how to apply it to the Nuri curriculum has been raised. Accordingly, this study defined early childhood AI education through literature research, identified the contents of AI education, and organized and operated it in the Nuri curriculum. As a results, AI education for children should be conducted for the purpose of cultivating digital capabilities based on computing thinking skills, and computers, the Internet, and programs were extracted as sub-elements of child AI education contents. Two approaches were proposed to incorporate this into the Nuri curriculum. The first is to set each of the three AI education contents as a life theme, select sub-factors accordingly, and plan and implement activities suitable for each sub-factors. The second is to develop and operate AI education contents at the level of sub-educational activities in accordance with the life theme of the existing Nuri curriculum. It is hoped that this study will consider the characteristics of early childhood education and be organized in the Nuri curriculum to realize the true meaning of early childhood AI education, and more research on AI play education programs according to the five areas of the Nuri curriculum.