SINGAPORE - Building Singapore’s own ChatGPT-like artificial intelligence (AI) model will be costly, but it is strategically vital to secure intelligence for its citizens to compete globally, said Mr Jensen Huang, chief executive of Nvidia, whose chips have powered the AI boom in recent times.

Singapore, the fourth-largest purchaser of Nvidia’s chips in the world, had on Dec 4 announced plans to build its own large language AI model similar to the ones built by the trio of American firms – OpenAI’s ChatGPT, Microsoft’s Bing and Google’s Bard.

The island state also has a game plan – dubbed National AI Strategy 2.0: AI for the Public Good for Singapore and the World – to lead global AI development and narratives. Among the actions to be taken is to secure a supply of specialised chips, which the industry believes will be from AI chip titan Nvidia, to power such computing workloads.

Speaking to local reporters at a roundtable at Shangri-La Singapore hotel on Dec 6, Mr Huang described the Republic’s move as pioneering, and said: “The question is, is Singapore’s digital intelligence important to the future of Singapore?”

The obvious affirmative answer is why many large software companies all over the world are jumping on building their own AI, he said. “What makes you is your intelligence. You can’t outsource it... you want to use your own data, manufacture your own intelligence.”

On why Western AI models may not be best suited for Singapore’s vision, he said: “You need to have data the country owns (to train the model). And there’s no reason to export the data so that you can import the intelligence. You should own it.”

Data used or generated by an AI model often contains information that provides competitive advantage. The ability to own and control the data is thus important for those investing in the use and development of AI.

The current AI frenzy over the development of large language models is expected to drive and sustain the demand for AI chips over the next 10 years.

“This is the beginning of a new technology cycle. We’re changing the way computers are built. We’re changing what a computer is used for,” said Mr Huang, noting that competition in the chip space will also become more intense with players such as Google, Amazon, Meta, IBM and Huawei.

The Taiwan-born executive arrived in Singapore on Dec 5 – his first visit here in 25 years – at the invitation of the Economic Development Board to explore new investment opportunities in AI. He will also be meeting Prime Minister Lee Hsien Loong.

“We have a supercomputer here already. We intend to build an even larger one,” Mr Huang said.

Nvidia, which leads in AI chip sales with more than 70 per cent market share, could also be potentially investing in “a significant iconic site for AI”, he said, noting that discussions are preliminary.

The discussions are part of the groundwork to develop the high-computing infrastructure needed to take Singapore forward in the global AI race. 

Singapore’s updated AI strategy – the inaugural one was unveiled in 2019 – has shifted the focus away from sector-specific projects to a longer-term view of nurturing talent, promoting a thriving AI industry and sustaining it with the needed infrastructure and research that ensures AI serves the public good.

The National Research Foundation will allocate $70 million to build expertise to develop Singapore’s large language AI model. This model will understand and incorporate the diverse cultures and languages of South-east Asia, and could eventually be used as the basis of various text-to-speech or text-to-image generative programmes. It will build on the early outcomes of AI Singapore’s South-east Asian Languages in One Network (Sea-Lion) model, an open-source system.

Nvidia is based in Santa Clara, California. In the three months ending October 2023, Singapore accounted for some 15 per cent of its revenue, trailing behind only the United States, Taiwan and China, according to a US regulatory filing. This translates to about US$2.7 billion (S$3.6 billion) worth of the purchases billed to Singapore for consumption by cloud services providers here as well as overseas.

Singapore is home to more than 70 operational data centres, accounting for 60 per cent of South-east Asia’s total data centre capacity.