AI agents function like personal assistants — they not only answer our questions but also perform various tasks, much like Jarvis, the assistant in the movie Iron Man. Jarvis is an advanced AI system that manages Iron Man’s mansion, armour and tech devices, while also providing strategic advice and avoiding risks. If AI agents are widely utilised, individuals could have multiple AI assistants, significantly boosting productivity and mitigating the challenges of slowing population growth.

Between 2016 and 2023, Singapore’s labour productivity grew by 2.4% annually. Prof Ho remarked that,

Some scenarios from science fiction films might gradually become reality by 2025.

“Have you seen the movie Big Hero 6? In the film, there’s an adorable healthcare companion named Baymax — chubby, huggable and endearing. Baymax is an AI system whose sole purpose is to collaborate with humans to promote their health, rather than replace them. One unique feature of Baymax is that he’s programmed never to harm anyone.”

Baymax exemplifies AI coexisting with humanity. Since AI systems are trained by humans, if the data used to train them is biased or unfair, the AI system itself would not be fair or just. Therefore, transparency in AI systems is critical. People need to understand the principles and logic behind these systems to trust and adopt them with ease of mind.

For Singapore, a small country striving to develop AI, credibility and responsibility are especially important. Data is the lifeblood of AI. Models rely on data to learn patterns and features that improve their accuracy and efficiency. Without sufficient high-quality data, it becomes challenging to build reliable AI systems.

Singapore’s characteristic is pragmatism. In the knowledge that it cannot compete with larger nations with vast domestic markets, Singapore has embraced an open and cooperative mindset, to expand the opportunities available.

AI scientist and venture capitalist Lee Kai-Fu, who was featured in Lianhe Zaobao’s Future 365 interview series in 2024, noted in his book AI Superpowers: China, Silicon Valley, and the New World Order that the world may be divided into two digital power blocs, led by the US and China, in what could resemble a new Cold War. These two economic giants are expected to dominate AI development due to their larger pools of venture capital, greater numbers of scientists, and most importantly, access to vast amounts of data. High-quality data enables the creation of superior AI products, whose users would generate even more data.

Since its independence, Singapore has continuously strived to break free from the limitations and destiny of being a small nation. The key to this lies in adopting a broad vision and acting swiftly and flexibly. In the realm of AI model development, Singapore refuses to confine itself to the large models developed by others. Instead, it recognises its unique position in Southeast Asia and aims to create its own large language model, similar to ChatGPT.

Prof Ho is especially excited when discussing SEA-LION. He explained in Mandarin during the interview:

“Existing large language models like ChatGPT use data primarily sourced from Western countries, and they often fail to capture the nuances of Southeast Asian languages. As a result, when we ask these models certain questions, they sometimes give irrelevant answers because their data doesn’t come from our local or Southeast Asian contexts.”

Currently, SEA-LION supports 11 commonly used languages in the Southeast Asian region: English, Mandarin, Burmese, Filipino, Khmer, Lao, Malay (including Indonesian and Malaysian), Tamil, Thai, Vietnamese and Sumatran.

While Singapore’s population is only about six million, the broader Southeast Asian region has nearly 700 million people speaking approximately 1,200 languages. Covering all these languages is undoubtedly a daunting task, but someone has to take the lead. Over time, this project is expected to become a critical AI resource for the region.

Prof Ho expressed satisfaction with SEA-LION’s progress over the past year. As an open-source project, it provides a platform for countries in the region. For instance, Indonesia and Thailand, two major regional powers, have used SEA-LION as the foundation for building their localised large language models, Sahabat and WangchanLION.

He explained that SEA-LION not only offers neighbouring countries a platform to develop their own large models but also benefits Singapore through collaboration. These partnerships have brought in a wealth of accurate and valuable data, significantly contributing to the improvement of SEA-LION.

Beyond receiving national-level support, SEA-LION has also attracted the attention of major tech companies like Google and Microsoft, which have begun recognising its potential. With the combined efforts of corporate and governmental partnerships, both the quantity and quality of SEA-LION’s data have been significantly enhanced. Prof Ho is confident that SEA-LION will become the most comprehensive large language model tailored to Southeast Asia.

“Singapore is a place that people generally trust, especially during times of global geopolitical division. If Singapore can serve as a bridge in Southeast Asia, it is partly because of our responsible management approach, which has earned praise.”

On his fourth day in office, US President Donald Trump signed an executive order related to AI, declaring the goal to “make America the world capital in artificial intelligence”. He also instructed an interagency task force to draft an “AI action plan” within 180 days to maintain and strengthen the US’s global AI dominance, aiming to boost economic competitiveness and national security.

Prof Ho said: “In the competition between the two major powers, the US and China, Singapore has always maintained a neutral stance. We welcome AI talent from around the world to come here and develop their careers. Here, a talented individual can collaborate with US experts, China experts, as well as Europeans and Indians.