Cloud & AI Innovation 2023' hosted by ITWorld and CIO Korea ended successfully. Held on March 29th at Lotte Hotel World Crystal Ballroom in Jamsil, this conference was attended by a total of 334 IT experts and focused on AI business strategy and multi-cloud utilization, which have become essential tasks for companies, under the theme of 'Cloud and AI that innovates future business.
Practical strategies Phone Number List and cases were introduced, such as ChatGPT, generative AI, AI-based cloud operation innovation, open source-based data architecture design, and cloud-native data analysis.
Errol Kuhlmeister, AI and data advisor at The AI Framework, former head of AI at H&M Group, served as the keynote speaker. “MIT Sloan’s 2019 research shows that most AI initiatives are not successful,” said Errol Kuhlmeister. Seven out of 10 companies said they had not achieved any results. The 2021 version of the same research report was quite different. 92% of large companies responded that they had realized profits from their AI investments. This is a huge change in just three years. AI is no longer a factor for securing competitiveness. “It is essential,” he emphasized.
Advisor Errol Kullmeister then pointed out, “AI is essential, but not many companies are doing it properly.” “The product and service development cycle in the IT industry is too long. Siled organizations and waterfall approaches are also not suitable for AI. There are too many managers. “The biggest challenge with AI is not algorithm development but organizational change.”
Errol Kuhlmeister, AI and data advisor at The AI Framework, former head of AI at H&M Group, served as the keynote speaker. ⓒFoundry
Advisor Errol Kuhlmeister advised applying AI to problems that can be solved most easily and quickly. “For example, H&M Group used AI in areas that have value and feasibility, that is, areas that can generate sales quickly. “We have created value by applying AI to end-of-season sales, etc., and it has become easier to request related investments,” he said. “In addition, a team should be formed where all stakeholders can cooperate and make decisions.”
Hyosung Information HPC Business Team consultant Kim Hyeong-seop, who was the second speaker, began his presentation by mentioning concerns in the field regarding AI. “What kind of work can it be applied to? What expertise do you need to know for AI business? What is the AI application process and considerations? What items are needed when introducing AI? Reference case? “This is a question many customers ask.”
Consultant Kim Hyeong-seop said, “The most important thing is that the types of AI model algorithms used in each industry are different. For example, the automotive and robotics sectors require high-performance computational resources (e.g. reinforcement learning, etc.), and the logistics and distribution sectors require relatively light computational tasks. Therefore, it is important to consider that the AI solutions needed are different depending on the industry and task. In addition, in order to have a smooth AI workflow from data source collection, data purification, data storage, data ops, and machine learning ops, users (e.g. developers, data analysts, data scientists, etc.) need to develop algorithms, and system administrators need the necessary resources. “We must be able to focus on distributing it in a timely manner.”
Consultant Kim Hyeong-seop then said that the company supports 'AI workflow for everyone', including pre-defined images, GPU partition virtualization, container-based GPU scaling, web browser-based AI model development environment, and Super Pod consisting of H100 and NV Link network. While introducing various services of Hyosung, he added that these products and services can be directly experienced at the DX Center.
In the expert session that followed, Dinodo APAC's Shanmuga Sundar Munni&D Data Architecture Director gave a presentation on the topic of 'Agile Logical Data Architecture for Innovative Enterprise Data Utilization'. He pointed out that the most important thing for a successful AI initiative is to input the right data, but many companies are unable to secure the necessary data because the data is scattered all over the place.
“At first, we imported and analyzed data from the mainframe,” said Shanmuga Sundar, director of Munni&D. Then, in the 90s, data warehouses came out, and data was copied and placed there. However, when data warehouses could not cope with various data types, data lakes appeared. Data was copied and placed into the data lake, and the number of such data lakes increased. “The result is a complex architecture.