생성형 AI의 적용 분야가 광범위해지는 가운데, 기업은 성공적인 생성형 AI 도입을 위해 각 기업의 목적에 맞는 클라우드 환경 구축의 필요성이 대두되고 있다. 삼성SDS는 맞춤형 서비스를 제공해 기업의 생성형 AI 자동화를 통한 생산성 향상을 지원한다.
.jpg)
▲Samsung SDS Vice President Kwon Young-jun
Steps for discovering application cases, deciding on an introduction method, and optimizing AI models
Samsung SDS provides integrated generative AI platform, cloud, and management services As the application areas of generative AI are expanding, companies are increasingly faced with the need to build a cloud environment that suits each company’s purpose in order to successfully introduce generative AI. Samsung SDS provides customized services to help companies improve productivity through generative AI automation.
Samsung SDS Vice President Kwon Young-jun explained the strategies needed for companies to adopt generative AI (Generative AI) at the Industrial Technology Association Breakfast Seminar held at the L Tower in Yangjae Station on the 14th, under the topic of how companies will utilize Gen AI (generative AI).
Since the launch of OpenAI's ChatGPT, most companies are expecting to increase productivity through the introduction of generative AI, but are hesitating to introduce it due to concerns about security, accuracy, and cost. In order to introduce generative AI, companies go through a series of processes such as discovering use cases for each industry, deciding on an introduction method, and learning and optimizing AI models based on company-specific criteria.
The first step for companies to adopt generative AI is to discover precedents. They analyze technology trends and review use cases to determine whether adoption is possible and establish a plan. After preliminary preparations are completed, the introduction of LLM (Large Language Model) and platform (CSP) is considered according to the conditions for introducing generative AI. At this time, internal corporate data is secured and the utilization is analyzed to determine the construction environment. The suitability of public cloud, private cloud, or hybrid cloud environment is determined.
Then, each company must apply and develop it to its business area. The work of 'fine-tuning' LLM exclusively for the company and continuously improving data to minimize hallucinations. LLM performance is also monitored for accuracy, reliability, and efficiency, and internal training is conducted to spread it.
Vice President Kwon said, “We need to establish a dedicated organization to promote the introduction of production-type AI within companies and provide specialized AI consulting in manufacturing, finance, healthcare, etc.” He added, “Samsung SDS provides LLMOps management services from the introduction stage to data-based utilization and service development, data operation management, and prompt engineering.”
Samsung SDS is Korea's only CSP (Cloud Service Provider) and MSP (Managed Service Provider). It provides integrated services that provide cloud and platform services and manage them. Samsung SDS's main services include: △Fabix, an enterprise generative AI service implementation and integrated management platform; △Brity Copilot, a generative AI business automation and collaboration solution; and △Samsung Cloud, which provides corporate data management, high-performance AI resources, and security.
'Fabrix' supports hyperautomation by utilizing internal corporate data. With Fabrix, companies can easily connect each company's business system and generative AI platform. Vice President Kwon said, “By linking generative AI to ERP operations, there is an effect of work efficiency and cost reduction through automation.” In addition, in SW engineering work, development efficiency can be improved by improving quality during coding work.
Vice President Kwon said, “After ChatGPT, AI trends will not stop at LLM, but innovation through AI hardware and devices based on LMM (Large Multimodal Model) is expected.” He also said, “Each company should focus on developing models and services specialized for each industry through internal data to differentiate by industry.”
He added, “In the manufacturing sector, too, a multimodal model combining visuals, sound, and language will likely be needed to increase safety and productivity.”