생성형 AI의 도입방안에 관한 연구 : 클라우드 방식과 온디바이스 방식에 대한 사고프로세스와 트리즈 원칙의 적용 (Research on Ways to Introduce Generative AI: Balanced Solution of Cloud-based AI and On-Device AI by Thinking Process and TRIZ's Principle)

서지정보

· 발행기관 : 한국인터넷전자상거래학회
· 수록지 정보 : 인터넷전자상거래연구 / 24권 / 6호 / 37 ~ 52페이지
· 저자명 : 강재정, 김근형

초록

This study is to examine the advantages and disadvantages of cloud-based AI and on-device AI, and to suggest ways to solve the conflicting problem between cloud-based AI and on-device AI for the companies considering introducing Generative AI through application of EC(evaporating cloud) of thinking process and separation principle of TRIZ. For this is to examine what are hidden or false assumptions and premises in the relationships of components in evaporating cloud diagram, and to suggest solutions through the breakdown of assumptions or premises by using the principle of separation of TRIZ, and various solutions are proposed according to space, time, condition and scale separation.
The guidelines in this study are as follows. First, it is the level of secret of personal or corporate internal information that required in the learning process of AI. If the level of secret for personal or corporate internal information is not required in the learning process, there is no problem even if you use cloud-based AI on Internet, but as the level of secret for personal information and confidential information increases, it is desirable for the company to operate and use on-device AI. Second, it is the level of data distribution for AI learning. If data for AI learning are distributed in many places and have to be collected in one place, Cloud-based AI is desirable. Third, in case there is high anxiety about cloud-based AI provided by third parties, it is desirable for company to purchase a foundation model, and to fit it appropriately, and use it exclusively on your purpose as a way to respond to data security risks. This study proposed various solutions for balanced operation of cloud-based AI and on-device AI, and showed that companies could secure information accuracy and security when introducing Generative AI.

영어초록

This study is to examine the advantages and disadvantages of cloud-based AI and on-device AI, and to suggest ways to solve the conflicting problem between cloud-based AI and on-device AI for the companies considering introducing Generative AI through application of EC(evaporating cloud) of thinking process and separation principle of TRIZ. For this is to examine what are hidden or false assumptions and premises in the relationships of components in evaporating cloud diagram, and to suggest solutions through the breakdown of assumptions or premises by using the principle of separation of TRIZ, and various solutions are proposed according to space, time, condition and scale separation.
The guidelines in this study are as follows. First, it is the level of secret of personal or corporate internal information that required in the learning process of AI. If the level of secret for personal or corporate internal information is not required in the learning process, there is no problem even if you use cloud-based AI on Internet, but as the level of secret for personal information and confidential information increases, it is desirable for the company to operate and use on-device AI. Second, it is the level of data distribution for AI learning. If data for AI learning are distributed in many places and have to be collected in one place, Cloud-based AI is desirable. Third, in case there is high anxiety about cloud-based AI provided by third parties, it is desirable for company to purchase a foundation model, and to fit it appropriately, and use it exclusively on your purpose as a way to respond to data security risks. This study proposed various solutions for balanced operation of cloud-based AI and on-device AI, and showed that companies could secure information accuracy and security when introducing Generative AI.

참고자료

· 없음