Samsung Electronics has reportedly initiated the full-scale development of an in-house Large Language Model (LLM) for internal purposes, following a recent internal data leak. The project, led by Samsung Research, has received substantial investments in terms of workforce and resources. All available GPU resources have been dedicated to this LLM development, a vital aspect of model training. Samsung aims to complete the initial version of its LLM within two months.
The main areas of focus for Samsung’s LLM solution will be document summarization, software development, and language translation. While the company hasn’t decided on public availability of the AI solution, internal use takes precedence.
The data leak incident resulting from the reliance on third-party generative AI solutions, including ChatGPT, prompted Samsung to prohibit their usage and embark on creating its own AI technology.
Samsung believes that deploying LLM-based AI solutions can significantly reduce the time required for software development and semiconductor design. Although Samsung explored partnerships with prominent companies like Google, Microsoft, Naver, and OpenAI, a top-level management meeting, including the presence of Lee Jae-yong, led to the decision of developing an in-house AI solution.
Several other South Korean firms, including Kakao, LG, Naver, and SKT, are also developing their own LLM-based generative AI solutions such as KoGPT, ExaOne, HyperCloverX, and A-Dot.
These solutions demonstrate human-like drawing and writing capabilities. However, challenges related to data accuracy and information privacy pose significant obstacles that need to be addressed during the ongoing development process.