B80m Thai AI Initiative Unveiled
Introduction to Thailand's New National AI Initiative
Thailand is taking significant steps towards establishing its own artificial intelligence (AI) infrastructure with the launch of a new large language model (LLM) project. This initiative aims to ensure AI sovereignty and support the country’s digital economy by developing a foundational AI model that deeply understands the Thai language, culture, and context.
The project, which has been allocated 80 million baht, is being led by the Ministry of Digital Economy and Society (DES) in collaboration with the Ministry of Higher Education, Science, Research and Innovation. The goal is to create a national AI infrastructure that can serve as a backbone for various sectors, from government agencies to startups.
The Vision Behind ThaiLLM
According to Patchara Anuntasilpa, the permanent secretary of the DES, digital technology and AI are no longer just support tools but have become fundamental components of future infrastructure. He emphasized the urgency of this project, noting that while traditional economic growth has slowed, the digital economy continues to grow steadily at around 3-4% annually.
"The real-world usage data generated from this massive rollout will be fed back into ThaiLLM to perfect the model," he said. This approach ensures that the model continuously improves based on actual user interactions and feedback.
Tiranee Achalakul, president of the Big Data Institute, highlighted the cost-effectiveness of having a national AI model. "Having a national model significantly reduces costs for government agencies and businesses, including startups, to build self-hosted applications as developers will not be burdened by token fees, allowing rapid, cost-effective scaling."
Focusing on Local Context
Unlike global tech giants that invest billions in AI development, ThaiLLM focuses on local context. It is designed to perform well when queried about Thai laws, culture, and specific localized knowledge. This makes it particularly useful for applications that require deep understanding of the Thai environment.
To achieve this, the project has established a National Data Bank that gathers data not only from web scraping but also from contributions by government bodies such as the National Research Council of Thailand, the Office of the Council of State, and the National Library.
Open-Licensed Models for Developers
ThaiLLM has developed 8B and 30B foundation models under an open-licence and open-weight approach. This allows developers to download and build specialized, fine-tuned models tailored to specific needs. The models are trained using public datasets and data contributed through collaborations across both public and private sectors, including legal, research, and news-related information within the Thai context.
The project is funded by the Digital Economy and Society Development Fund with a budget of 80 million baht in fiscal 2024. Nearly 8,000 developers are already testing and using the platform's application programming interfaces and chat sandboxes.
Real-World Applications
Real-world applications are already live, including a medical screening chatbot that provides preliminary health advice and a tourism bot. These applications demonstrate the practical benefits of the ThaiLLM project.
"This is state infrastructure preparing to be passed on for people to build upon for free," said Ms Tiranee. "Ultimately, it should be able to advance faster because we have our own foundation."
Ensuring Data Security and Cost Efficiency
By using the LANTA supercomputer at the National Science and Technology Development Agency for local processing, Thailand can significantly reduce cloud costs while ensuring sensitive national data remains securely within the country, according to Punpermsak Aruni, deputy permanent secretary of the higher education ministry.
The ministry is also mobilizing AI researchers and specialists from 150 universities and six research institutions nationwide to refine datasets and strengthen the model's understanding of complex Thai contexts, including government systems and legal frameworks.
Advancing AI Workforce Development
In addition to building the AI infrastructure, the project is accelerating AI workforce development through new curricula and training programs. These initiatives aim to prepare talent for the growing AI infrastructure and ensure that Thailand remains competitive in the global AI landscape.
