Singapore becoming AI hub with commercial models in local languages


Article Contents

In This Article

The island nation of Singapore is eyeing 2024 as the year to scale its Southeast Asian-focused artificial intelligence (AI) models for commercial use and become a global hub for the technology. 

A government-led initiative called AI Singapore is leading the way in developing the country’s own AI models, particularly those that cater to local populations.

SEA-LION is an open-source ChatGPT-like large language model (LLM) built by AI Singapore to represent the languages and cultures of southeast Asia and is set to be the first in a series of SEA-LION LLMs. The model has been trained on data in 11 languages, such as Vietnamese, Thai, and Bahasa Indonesia, which it plans to have ready in 2024.

Dr. Leslie Teo, senior director of AI products at AI Singapore, told Cointelegraph that they plan to release a better instruction-tuned model in the next few weeks.

“This will come in a version that will be commercially permissible as well as one that is not licensed for commercial use due to the data we used,” he said.

“We are building a larger corpus for pre-training and hope to hit 2-3 trillion soon. Once that is done, we will further develop SEA-LION to 30-50B parameters this year, and extend SEA-LION into a multimodal speech-text model.”

SEA-LION will help countries in Southeast Asia lessen their reliance on models built and developed in the West for the English language. 

874194e5 101e 4cbb 9b24 f762d2bddb92
Prompt given to SEA-LION in local language and three other LLMs. Source: AI Singapore
55dfbd94 8617 4786 a777 859cb61248e5
Response from LLMs with SEA-LION being the quickest. Source: AI Singapore
b40f46e8 b4d3 4394 b3c2 d61594ccb21c
Description of answer quality from AI Singapore saying SEA-LION was the fastest and most competent followed

Singapore is well known in the emerging tech spaces for being a leading country and hub for cryptocurrency and blockchain technology development. Recently, a Google Cloud executive told CNBC that it has a “very high” potential to add AI to that list due to it having an environment that fosters innovation. 

Teo echoed that sentiment, saying that AI Singapore plans to expand its network of collaborators:

“… to build a vibrant community that will build better datasets, evaluation metrics, and models to shape and contribute to the continued development of SEA-LION for Southeast Asia.”

Already, Teo said they have worked with Amazon Web Services (AWS) for infrastructure and Google Research for data. Additionally, he mentioned partnerships with communities such as SEACrowd to accelerate the creation of a diverse data corpus in native languages. 

Related: Google’s 25M Euro AI skill boost for Europe unveiled

Through these collaborations, AI Singapore aims to develop SEA-LION’s capabilities and push its adoption by various organizations across the region.  He said:

“We believe that the SEA-LION model will set new benchmarks in language learning and understanding, paving the way for more advanced and efficient natural language processing models in the future.”

Teo added that SEA-LION is open and can be used by companies seeking to do business and reach the Southeast Asia region.

According to data from KPMG, AI fintech funding in Singapore hit $333.13 million during the last half of 2023, which is a 77% increase from the first half of the year. This made a total of $481.21 million across 24 deals in the AI sector in Singapore last year.

At the end of 2023, Singapore released its National AI Strategy 2.0, in which it revealed plans for 15,000 AI experts, boosting government competency, building a smart nation, and increasing computing capacity.

Magazine: ChatGPT trigger happy with nukes, SEGA’s 80s AI, TAO up 90%: AI Eye