0:00
0:00
NECTEC Unveils Pathumma LLM: Thailand’s Open AI Model for Localized and Multi-Modal Applications
The National Electronics and Computer Technology Center develops an open Thai language model with text, audio, and vision capabilities, aiming to progress toward Agentic AI.
The National Electronics and Computer Technology Center (NECTEC) under the National Science and Technology Development Agency has developed Pathumma LLM, a Thai Large Language Model designed to support local language processing and context-specific applications.
Developed as an open model, Pathumma LLM aims to provide flexibility and data confidentiality for sectors such as banking, healthcare, and judicial services, as well as to serve government applications like drafting official documents and handling regional dialects.
According to Sarawut Kongyang, a researcher with NECTEC's Artificial Intelligence Research Group, large language models are typically divided into closed models, which restrict public access for further development, and open models that are available for further enhancement.
Pathumma LLM is offered in three formats: a web application for general users, an API for integration by developers, and a downloadable model for further development.
Pathumma LLM features Multi-Modal Generative AI capabilities, supporting three types of data processing.
The Text LLM component is optimized for handling text queries and logical reasoning, making it suitable for enterprise applications such as tax-related chatbots for the Revenue Department.
The Audio LLM processes audio data by transcribing speech in Thai and English, generating sound descriptions, and detecting speaker attributes.
The Vision LLM is designed to analyze images by generating descriptions, extracting text, and answering queries related to visual content.
The research team plans to enhance the model's efficiency by collaborating with public and private sector partners to develop a larger foundation model, with implementation expected to begin by February 2025. They intend to request publicly available data from Thai government agencies and private organizations to build a robust knowledge base, thereby strengthening the model’s expertise in Thai language and context.
Future development goals include advancing Pathumma LLM toward Agentic AI, which is defined as artificial intelligence capable of autonomous reasoning and decision-making.
The team envisions applications such as personal AI assistants that can analyze market trends and execute predefined tasks, or smart chatbots that can recommend services and track task statuses.
The project aligns with global trends in AI technology and aims to contribute to enhanced government service efficiency through AI-driven solutions.
Developed as an open model, Pathumma LLM aims to provide flexibility and data confidentiality for sectors such as banking, healthcare, and judicial services, as well as to serve government applications like drafting official documents and handling regional dialects.
According to Sarawut Kongyang, a researcher with NECTEC's Artificial Intelligence Research Group, large language models are typically divided into closed models, which restrict public access for further development, and open models that are available for further enhancement.
Pathumma LLM is offered in three formats: a web application for general users, an API for integration by developers, and a downloadable model for further development.
Pathumma LLM features Multi-Modal Generative AI capabilities, supporting three types of data processing.
The Text LLM component is optimized for handling text queries and logical reasoning, making it suitable for enterprise applications such as tax-related chatbots for the Revenue Department.
The Audio LLM processes audio data by transcribing speech in Thai and English, generating sound descriptions, and detecting speaker attributes.
The Vision LLM is designed to analyze images by generating descriptions, extracting text, and answering queries related to visual content.
The research team plans to enhance the model's efficiency by collaborating with public and private sector partners to develop a larger foundation model, with implementation expected to begin by February 2025. They intend to request publicly available data from Thai government agencies and private organizations to build a robust knowledge base, thereby strengthening the model’s expertise in Thai language and context.
Future development goals include advancing Pathumma LLM toward Agentic AI, which is defined as artificial intelligence capable of autonomous reasoning and decision-making.
The team envisions applications such as personal AI assistants that can analyze market trends and execute predefined tasks, or smart chatbots that can recommend services and track task statuses.
The project aligns with global trends in AI technology and aims to contribute to enhanced government service efficiency through AI-driven solutions.