|
Listen to this story
|
When countries are working towards building their regional LLMs, one more joins the race from the south-east side. Singapore’s Infocomm Media Development Authority (IMDA), alongside AI Singapore and the Agency for Science, Technology, and Research, has unveiled the National Multimodal LLM Program (NMLP), an AI initiative with a budget of S$70 million ($52.3 million). The program, funded by Singapore’s National Research Foundation, is set to enhance the nation’s research and engineering capabilities in multi-modal large language (LLM) models.
The NMLP aims to develop a base model with regional context. This model will not only cater to Singapore’s unique linguistic characteristics but also address the diverse culture and languages prevalent in the Southeast Asian region.
The initiative is a part of Singapore’s Research, Innovation, and Enterprise 2025 plan and will also support Singapore’s national AI strategy 2.0. The overarching goal of this initiative is to cultivate a more inclusive and regionally relevant AI landscape. This move is a first-of-its kind in South East Asia, which will help strengthen Singapore’s position in the AI research space.
Singapore is not starting from scratch to create the first Large Language Model (LLM) in the region. Instead, it will build upon the existing work of AI Singapore’s Sea-Lion model (Southeast Asian Languages In One Network), an open-source LLM that reflects the cultural nuances of Southeast Asia.
Sea-Lion is designed to be smaller, more flexible, and faster than commonly used LLMs. It also provides a cost-effective and efficient option for organizations with budget constraints and throughput limitations looking to incorporate AI into their workflows.
Demographic-Specific Models
The recent growth of regional language models has been on the rise. UAE has been on the forefront with Core 42’s Arabic language model Jais 13B and 30B. China has also followed a similar route with DeepSeek, a bilingual LLM (67 billion parameter model) which was released recently and is available in both English and Chinese.
