[K-Construction, A New Leap] Hyundai Engineering
Co-developed with AI Startup 'Zenti'
High Response Reliability... Increased Work Efficiency
Chat File and Bid Guide Comparative Analysis Service Also Available
Hyundai Engineering unveiled a specialized large language model for the plant and construction sectors at its technology conference 'AI REDAY'. (Photo by Hyundai Construction)
"Tell me how to calculate the required piping quantity for construction." (Hyundai Engineering employee)
"I will explain the piping quantity calculation method by material and welding type." (Plant and Construction specialized LLM)
Hyundai Engineering has completed the development of the world's first conversational artificial intelligence (AI) optimized for plant and construction tasks. It is expected to significantly improve employees' work efficiency through practical support such as information retrieval and document generation.
According to Hyundai Engineering on the 11th, the company unveiled its self-developed large language model (LLM) specialized in the plant and construction field at its technology conference 'AI READY' held last October. An LLM is an AI model that learns from large-scale text to process and generate language similarly to humans. ChatGPT, developed by OpenAI, is a representative example.
To develop the LLM, Hyundai Engineering has been collaborating with AI R&D startup 'Zenti' since last year. Hyundai Engineering was responsible for providing plant and construction domain data and knowledge information, while Zenti handled AI language model research and development. The goal was to develop a specialized program recognizing the limitations of general services like ChatGPT. ChatGPT has restrictions for industry use due to security issues preventing input of internal company data, as well as problems such as generating false information, lack of up-to-date information and expert knowledge, and cost concerns.
As a result, they developed a foundation model trained on a vast plant construction dataset consisting of 16.5 billion corpus tokens. By training on specialized engineering materials and refined internal data, hallucination was reduced and answer reliability was improved. The application (app) to utilize the completed LLM in work will be jointly developed by both companies.
Hyundai Engineering expects this program to greatly enhance productivity and work efficiency through features such as reduced information search time, automatic document generation, risk analysis, technical decision-making, and establishment of employee training systems.
They are also focusing on developing additional services using this model. These include a chat-file service that allows searching, summarizing, and translating vast internal technical documents via Q&A, and a service that compares and reviews bid invitation (ITB) items based on past cases, legal provisions, and standard contract conditions (FIDIC). Once the program development is complete, employees will be able to obtain refined data and documents simply by entering basic questions.
A Hyundai Engineering official emphasized, "We will continue to promote digital transformation and IT infrastructure innovation to improve company-wide work efficiency."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
