TL;DR:
- Singapore is developing a unique large language model (LLM) tailored for South-east Asian cultures and languages.
- The LLM could revolutionize text-to-speech and text-to-image generative programs.
- This initiative addresses the gap left by Western-dominated LLMs.
- Collaborators include IMDA, AI Singapore, and A*Star, funded by the National Research Foundation.
- Aligned with Singapore’s National AI Strategy 2.0, it aims to nurture local AI talent and foster AI industry growth.
- The initiative enhances AI governance and promotes cross-border industry collaboration.
- Building on Sea-Lion, it will create a 30-50 billion parameter multimodal speech-text model.
- Sea-Lion covers 11 regional languages and enables contextual understanding.
- Singapore’s investment solidifies its position as a global AI hub.
Main AI News:
In the rapidly evolving landscape of artificial intelligence (AI), Singapore is taking a groundbreaking step forward. The nation is spearheading the development of a sophisticated large language model (LLM) that is uniquely tailored to understand and generate human language with a focus on the diverse cultures and languages prevalent in South-east Asia. This ambitious endeavor marks a significant stride towards bridging the existing gap in the AI landscape, which predominantly features models rooted in Western cultures, values, and norms, a world apart from Singapore and its neighboring nations.
The potential applications of this South-east Asian-centric LLM are boundless. It may serve as the foundation for various text-to-speech or text-to-image generative programs, ushering in a new era of AI-driven innovation. Renowned LLMs like OpenAI’s ChatGPT, Microsoft’s Bing, and Google’s Bard have already demonstrated the transformative power of such models.
The initiative, known as the National Multimodel LLM Programme, is a collaborative effort between the Infocomm Media Development Authority (IMDA), AI Singapore, and the Agency for Science, Technology and Research (A*Star). With a generous $70 million in funding from the National Research Foundation, this program is set to boost Singapore’s prowess in AI research and innovation significantly.
This strategic move aligns perfectly with Singapore’s National AI Strategy 2.0, a visionary roadmap unveiled by Deputy Prime Minister Lawrence Wong at the recent Singapore Conference on AI. The LLM project’s primary objective is to nurture a pool of skilled AI talent within Singapore. It accomplishes this by providing financial support and granting access to cutting-edge computing resources for local researchers and engineers. Simultaneously, it strives to cultivate a thriving AI industry that can harness LLM-enabled solutions, thus enhancing productivity and creating novel opportunities for businesses.
Moreover, this initiative will play a pivotal role in establishing Singapore as a trusted hub for AI utilization. By delving deeper into the mechanics of LLMs, it promises to facilitate a more profound understanding of AI operations, thus fostering research in AI governance. Dr. Ong Chen Hui, Assistant Chief Executive of IMDA’s Biztech group, affirms this national effort as a testament to Singapore’s commitment to becoming a global AI hub. He emphasizes the pivotal role of language as an enabler for collaboration and underscores the nation’s ambition to stimulate cross-border industry cooperation, driving the next wave of AI innovation across South-east Asia.
This endeavor builds upon the early successes of AI Singapore’s South-east Asian Languages in One Network (Sea-Lion) model. Sea-Lion is an open-source LLM designed to be more reflective of the region’s cultural contexts and linguistic nuances. What sets it apart is its smaller size, enhanced speed, and flexibility when compared to commonly used LLMs available today. It offers a cost-effective and efficient alternative to existing options.
The forthcoming national initiative aims to further refine Sea-Lion, increasing its parameter count to an impressive 30 billion to 50 billion and expanding it into a multimodal speech-text model. While some well-known LLMs are substantially larger—such as GPT-4 with approximately 1,700 billion parameters and Llama-2 with 70 billion parameters—Sea-Lion’s emphasis on regional relevance and efficiency is its defining characteristic.
Sea-Lion has been meticulously trained on 11 languages spoken in the South-east Asian region, including English, Chinese, Indonesian, Malay, Thai, and Vietnamese. The result is a versatile LLM that empowers users to comprehend the nuanced context and values associated with the diverse cultures and languages of South-east Asia. It excels in managing the complexities of language-switching in a multilingual environment like Singapore.
Conclusion:
Singapore’s $70 million AI initiative signifies a pivotal moment in the South-east Asian AI market. By developing a region-specific LLM, Singapore is not only addressing a crucial gap in AI representation but also positioning itself as a leader in AI innovation. This initiative promises to unlock new opportunities for businesses, nurture local talent, and foster cross-border collaboration, making Singapore a hub for AI development and deployment in the region.
