[Courtesy of NCSoft]
Large language models (LLMs) are widely being developed by various companies like OpenAI, Google, Microsoft, and other technology companies. ChatGPT is one of the most famous LLM-based chatbot services. An LLM is normally based on about one billion parameters, variables present in the model. ChatGPT is known to have some 100 billion parameters.
South Korean companies are also developing LLMs to catch up with the global tech trend. Web service giants Naver and Kakao, and the country's biggest telecom service provider KT are preparing their own LLM-based services. An LLM can be transfused into any solution like chatbots, interactive video content services, and web search solutions to offer a feeling that users are actually interacting with service solutions in human-like verbal conversations.
According to NCSoft, the company will introduce a self-developed LLM that has a smaller number of parameters for in-house use this month. A full LLM will be adopted later this year. The game publisher said that its full AI language model will have some 175 billion parameters. The smaller version would have some seven billion parameters.
Along with the in-house LLM, NCSoft will also release a "digital human" solution. The sample of the digital human, modeled after the company CEO Kim Taek-jin was showcased at the Game Developers Conference (GDC) 2023 in March 2023. The digital human, which enables developers to create detailed models of real-like artificial humans, will have the ability to better communicate with real humans if it is applied with an LLM.
"We have focused on the development of AI technologies based on ideas that games combined with AI can offer both the company and customers new values," said Lee Jang-wook, NCSoft's head of investor relations during the conference call in May.
reporter
Park Sae-jin
swatchsjp@ajunews.com
