Thailand is accelerating the development of the Open ThaiGPT project, hoping the country can avoid heavy reliance on foreign technologies in the area of artificial intelligence (AI).
Thailand needs to have its own version of the generative pre-trained transformer (GPT) model to support local needs by building a local ecosystem, Chai Wutiwiwatchai, executive director of the National Electronics and Computer Technology Center (Nectec), told the Bangkok Post.
GPT, a type of large language model (LLM) technology, can perform a broad range of tasks from creating original content to summarising text and writing code.
The country needs to have its own tech capabilities to minimise risks from a heavy dependence on foreign technologies, he said, as foreign tech companies could increase their service fees if Thailand lacks credible alternatives.
Nectec teamed up with three organisations in April 2023 to launch the Open ThaiGPT project, which is a next-generation form of AI using the Thai language.
The three organisations partnering Nectec are the Artificial Intelligence Entrepreneur Association of Thailand, the Artificial Intelligence of Thailand (AIAT) and the NSTDA Supercomputer Center.
The project offers a customisable chat-based assistant with state of the art AI technology and the ability to understand Thai language.
The Digital Economy and Society Ministry and the Higher Education, Science, Research and Innovation Ministry have supported the development of specific use cases of AI chatbots in several areas, notably tourism and medical, and in the form of a GPT-assisted system of document search for searching Thai-language R&D papers.
Thepchai Supnithi, director of Nectec's AI Research Group and vice-president of the AIAT, said progress has been made with the Open ThaiGPT project, which is developing a Thai chatbot system with capabilities equivalent to OpenAI's ChatGPT, as well as being able to connect to external systems and retrieve data flexibly. It's easily expandable, customisable and has been developed into free open-source software.
The Open ThaiGPT system now has 13 billion parameters, up from 7 billion parameters, and there are plans for it to reach 70 billion parameters, he said.
Thailand is similar to many countries that are in a good position to have their own language GPT model such as Japan, South Korea, Norway and other European countries. Singapore is developing Southeast Asia's inaugural LLM ecosystem, tailored to the diverse cultures and languages of the region, he added.
Mr Thepchai said the Open ThaiGPT platform is able to serve data management for specific business uses. Businesses can leverage it to support their work, such as searching for information, and it could be a foundation for foreign tech and business firms, including the Government Big Data Institute, to use the framework. However, all would need to ensure the use of AI was responsible and explainable.
GPT models will be useful in government services and in areas of the private sector such as banking and industry, he said.
According to Reuters, OpenAI launched its GPT Store as a marketplace for personalised AI applications. The store is a place for users to discover and build GPTs, or AI customised for tasks such as teaching math or designing stickers.
Mr Chai said a GPT store would enable more revenue opportunities for developers as low-hanging fruit for custom GPTs, but they need to collaborate with linguists to build massive databases.
He said Thailand might consider having its own GPT store to provide custom GPT solutions in the country to ensure security, privacy, standardisation and transparency for developers to support specific business uses. The custom GPTs should cover three areas: tourism, medical and e-government.
Panutat Tejasen, chief executive of AI startup ThaiGPT Co, said OpenAI's GPT store will enable developers to develop custom GPTs to reach consumers as a new source of revenue opportunities.
For example, super users such as English teachers can use ChatGPT as a foundation to create 'English-Thai courseware' to support students' learning.
Mr Chai also considers proposing guidelines on ethics to ensure AI services guarantee LLM data transparency, explainable AI, and no bias. The guidelines must apply to GPT models for government-related tasks and services.
He said AI can be used to identify the bank accounts of those suspected of carrying out scams by analysing account movements. This helps authorities to suppress mule accounts.
In a related matter, Thailand is scheduled to host the 62nd annual meeting of the Association for Computational Linguistics this year in Bangkok on Aug 11-16 at CentralWorld. The event will feature an exhibition of global tech firms involved with the AI movement. This offers an opportunity for Thailand to gain global recognition for its AI capabilities and as a destination.
"This will help speed up AI project development. A two-year project will be reduced to 18 months via collaboration among experts," Mr Thepchai noted.