DTGO partners with leading AI software company SenseTime to launch “DTLM”, a new trilingual LLM for Thai, Chinese, and English

 
DTGO Corporation Limited, the parent group of MQDC, has partnered with SenseTime, a leading AI software company headquartered in Hong Kong, to develop DTLM (DTGO Large Language Model).
DTLM is a large language model (LLM) that leverages AI development expertise and ownership of DTGO and SenseTime’s big data for proficiency in three languages: Thai, Chinese, and English. The model is designed to efficiently meet multilingual requirements across these three languages, offering high-performance capabilities in each while understanding the local Thai language and culture.
Mr. Kittikun Potivanakul, Chief Technology Officer of DTGO, said that DTLM, DTGO’s first Thai LLM product, has been developed to accurately and thoroughly answer questions in Thai. It has achieved higher test results on the MMLU (Massive Multitask Language Understanding) benchmark than other Thai LLMs.
Through the collaboration with SenseTime, DTLM is an LLM that is built upon the company’s leading SenseNova Large Model Series. At the 2024 World Artificial Intelligence Conference (WAIC) in Shanghai last week, SenseTime unveiled its upgraded SenseNova 5.5 Large Model. Leveraging SenseNova 5.5’s significantly enhanced language capabilities, extensive knowledge database, and high information density, DTLM is able to engage in natural and responsive real-time conversations.
DTLM is the world’s first model to work efficiently in these three languages. Developed using machine learning technology that doesn’t rely solely on translation, DTLM provides an authentically local generative AI experience, including reading texts, asking questions, and responding naturally and quickly in real-time.
By empowering Thai customers with this innovative AI solution, DTLM aims to drive growth across various sectors and further enrich Thailand’s AI ecosystem, opening new opportunities and possibilities.
DTLM will initially focus on corporate clients, offering them cost-effective usage, helping control expenses efficiently, and incorporating a digital trust system for clients who require a highly dependable model.
DTLM can be applied in various ways, especially in vertical industries where internal organizational data can be used in conjunction with LLM to effectively answer questions about internal knowledge. It can also be used to develop services and external applications, such as in university libraries.
“We won’t stop here,” said Mr. Kittikun Potivanakul. “The world of technology is advancing rapidly, and we must develop to keep pace with industry advancements. In the near future, the DTLM model will be able to support usage for the general public as well.
“DTGO established Quinnnova as a subsidiary because we saw the potential for IoT and AI development, as well as opportunities to develop it further into one of the high-potential businesses of the future.”
Professor Lin Da Hua, Co-Founder, Executive Director, Chief Scientist for AI Infrastructure and Large Model of SenseTime, said, “We are excited to introduce DTLM, our first large language model built upon the leading SenseNova Large Model Series and tailored for the Thai community. Coupled with DTGO’s deep expertise of the local landscape, we hope that DTLM’s ability to understand the Thai language and culture will better address the diverse needs of Thai customers and businesses. Together, we will continue to drive innovation with DTGO and deliver cutting-edge AI solutions for our clients in Thailand and Southeast Asia that are effective, cost-efficient, and sustainable.”
DTGO is active across various sectors and in societal contribution. The group’s main business is MQDC, a leading real estate developer. DTGO emphasizes the importance of implementing leading technologies and innovations in all aspects of its operations, whether under the group or its subsidiaries.
The post DTGO partners with leading AI software company SenseTime to launch “DTLM”, a new trilingual LLM for Thai, Chinese, and English appeared first on HIPTHER Alerts.