Samsung starts developing a Large Language Model for internal use

Since the onset of ChatGPT, various tech companies are trying hard to lead the AI race. Most of them have responded to the surge in AI by developing their own AI models. Samsung is one of these companies. the South Korean tech company faced some data breaches with ChatGPT. For this reason, the company restricted its employees from using ChatGPT. Back then, the firm announced that it will create its own Large Language Model.

As of now, the recent pieces of information suggest that the company has begun the development of its own LLM for internal use. Reportedly, the company has started working on the AI model by this month. The work is conducted under the supervision of Samsung Research. Reports indicate that the company has utilized a huge workforce as well as investment for this project. LLM training requires GPUs. According to some officials, the company has used most of its GPU resources for LLM development. It is claimed by the company that the development of the first version of LLM will complete in two months.  

Samsung’s LLM-based AI solution will offer faster document summarizing and software development

Samsung intends to utilize the services of its LLM for language translations, document summarization, and software development. There is no word whether the AI solution by Samsung will be available to general consumers or not. According to reports, Samsung believes that these AI technologies can significantly save the time required for semiconductor design and software development.

The business was in discussions with a number of businesses, including Google, Microsoft, Naver, and OpenAI. However, it was ultimately decided that Samsung should create its own AI solution during the company’s top-level management meeting. Well, this is not just Samsung several other South Korean firms are also creating their own LLM-based AI solutions. These include LG’s ExaOne, Naver’s HyperCloverX, Kakao’s KoGPT, and SKT’s A-Dot. Most of these LLMs are designed to draw and write like humans. However, the two largest obstacles at the moment are data accuracy and information privacy.