LG CNS announced on the 31st that it will unveil a large language model (LLM) optimized for code-generating artificial intelligence (AI).
Previously, in June last year, LG CNS developed 'AI Coding,' a code-generating AI service based on ChatGPT, which has been utilized not only for internal system operations but also in customer system integration (SI) and system management (SM) projects.
LG CNS has also independently developed the LLM that acts as the brain of AI Coding, enabling companies sensitive to data leakage, such as financial and securities firms, to use it. Customers can choose between LG CNS's LLM or OpenAI's GPT according to their respective business environments to utilize AI Coding.
LG CNS implemented an LLM optimized for coding by leveraging various open-source LLMs, including Meta's CodeLlama. Customers can install LG CNS's LLM on their existing infrastructure, such as owned servers or cloud environments, and connect AI Coding as a plugin that adds additional functions to existing systems. Since customers use their own infrastructure, concerns about source code and personal information leakage can be resolved.
LG CNS also provides consulting to configure optimal prompts for customers who wish to adopt the newly released LLM and AI Coding. Prompts optimize questions by incorporating know-how to improve the accuracy of LLM responses. Generative AI experts, including LG CNS's prompt engineers, pre-configure prompts containing customer information such as coding standards, system development environments, and performance criteria, enabling customers to immediately utilize AI Coding.
LG CNS's AI Coding supports developers' coding tasks when building systems, applications, and services. It consists of an AI Coding Orchestrator and prompts. The Orchestrator is a tool that links the LLM to respond to users based on customer data. It also includes a function to prevent information leakage, which is the biggest issue in using generative AI services. For example, if a developer at a financial institution accidentally inputs personal information such as a customer's name, date of birth, or asset size, the Orchestrator detects and masks this data, processing it as pseudonymous information to prevent external leakage.
By utilizing AI Coding, developers can perform various tasks such as code conversion, code generation, code recommendation, and code quality inspection according to their requirements. Code conversion transforms legacy code into code compatible with the current system version. Code generation automatically creates code by analyzing the context of existing code based on natural language commands from developers.
Additionally, code recommendation allows developers to receive suggestions for the next lines of code during the writing process, and code quality inspection detects parts that cause errors and lets developers review automatically corrected code. LG CNS is preparing to advance AI Coding by incorporating coding expertise accumulated over more than 30 years of digital transformation (DX) across various industries such as finance, logistics, and manufacturing.
Kim Seonjeong, Chief Technology Officer (CTO) and Executive Vice President of LG CNS, stated, "We have secured an LLM optimized for AI Coding that can be installed and used in the infrastructure environments customers possess. Through AI Coding, we aim to enable customers to revolutionize the productivity of digital service development."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
